SPECIES TURNOVER AND RICHNESS OF AQUATIC COMMUNITIES IN NORTH TEMPERATE LAKES

by

SHELLEY ELIZABETH ARNOTT

A dissertation submitted in partial fulfillment

of the requirements for the degree of

Doctor of Philosophy

(Zoology)

at the

UNIVERSITY OF WISCONSIN-MADISON

1998 TABLE OF CONTENTS

Abstract ii

Acknowledgments iv

Preface 1

Chapter 1: 9

Crustacean zooplankton species richness: single- and multiple-year estimates.

Chapter 2: 45

Inter-annual variability and species turnover of crustacean zooplankton in Shield

lakes.

Chapter 3: 86

Long-term species turnover and richness estimates: a comparison among aquatic

organisms.

Chapter 4: 146

Lakes as islands: biodiversity, invasion, and extinction.

Summary of Thesis 173 II ABSTRACT

SPECIES TURNOVER AND RICHNESS OF AQUATIC COMMUNITIES IN NORTH

TEMPERATE LAKES

SHELLEY ELIZABETH ARNOTT

Under the supervision of Professor John J. Magnuson

At the University of Wisconsin- Madison

I estimated annual species turnover rates for three groups of aquatic organisms in relatively undisturbed north temperate lakes. Apparent turnover rates (i.e. measured turnover rates) were high, averaging 18% for phytoplankton, 16% for zooplankton, and 20% for fishes. Based on life history characteristics and dispersal abilities, I expected phytoplankton to have higher turnover rates than zooplankton, which would have higher turnover rates than fishes. Results were contrary to my expectations; apparent turnover was high and similar for each of the taxonomic groups. Comparison of apparent turnover rates, however, was problematic because sampling error could account for much of the apparent turnover.

Because the turnover that could be attributed to sampling error was so high, it should be taken into consideration when assessing species turnover. I have developed a new and unique method for quantifying potential sampling error in which I calculate the species turnover that could be attributed to failing to detect species that were present but at low abundance. Sampling error was influenced by sampling intensity. When sampling effort was increased, the proportion of turnover attributable to sampling error decreased. iii The result of high apparent turnover is that biodiversity assessments are dependent on the extent of the sampling programme. I assessed crustacean zooplankton richness in eight lakes at different temporal and spatial scales, using three methods of estimation: cumulative, asymptotic, and Chao's index. Percent species detected increased with the number of spatial, intra-annual, or inter-annual samples taken. Single samples detected only 50% of the annual species pool and 33% of the total estimated species pool. This suggests that previous estimates of zooplankton richness, based on single samples in individual lakes, were seriously underestimated. Single-year richness estimates provided poor predictions of multiple-year richness. The relationship between richness and environmental variables was dependent on the method of estimation and the number of samples used. I conclude that richness should be treated as an 'index' rather than an absolute and sampling efforts should be standardized.

Approved: iv ACKNOWLEDGEMENTS

I am indebted to many people who have contributed to this research over the past five years. Some have contributed directly through field and technical assistance whereas others have contributed by sharing their ideas and enthusiasm about science. Equally important are those who shared their friendship, provided support, and patience throughout this endeavor.

By no means is this thesis exclusively my own effort. It is a product of many people, including co-authors, advisors, mentors, colleagues, friends, and family.

I have had the tremendous good fortune to have several advisors; John Magnuson,

Tom Frost, and Norman Yan. John was my official advisor and mentor. His enthusiasm, positive approach, and excitement for the complexities and unanswered questions in science were a continuous source of inspiration. He guided me through my graduate studies with pep talks, discussions about how to be a scientist with diverse interests, and advise about teaching, dealing with difficult situations, and developing practical skills associated with being a scientist. Through this, and by John's example, I learned the importance of projecting self-confidence while still being able to laugh at yourself, the necessity of directness, the value of taking the time to observe wonders around you, and the importance of having a good heart.

Tom Frost has been an active mentor through both my master's and dissertation work.

He always seemed to have a knack for asking me the tough questions that I hadn't considered and was able to keep me grounded when my ideas became too airy-fairy. His encouragement and support were greatly appreciated.

Sometimes seemingly insignificant events lead to important changes in the direction of your work. I met Norm Yan because I had the 'nerve' to phone him and ask for all his v data. From that moment, Norm became a very active and important collaborator. I have learned much from Norm during the course of my dissertation. He has given me advice on writing, illustrated the importance of integrity, directness, and friendship in science. Norm has been an important mentor during the course of this research and continues to be a role model and inspiration.

I have also had the good fortune to interact/collaborate with several outstanding scientists. Dave Findlay, Alex Saiki, and Mike Patterson, from the Freshwater Institute in

Winnipeg have provided data, discussed analyses, reviewed papers, and helped shape and refine this thesis. Tony Ives, Stephen Carpenter, Diana Padilla, Stanley Dodson, and Eric

Triplett were valued members of my committee, contributing ideas and insights to some difficult problems. Advice and wisdom from Jim Kitchell was always appreciated.

Now comes the difficult part. How do I begin to acknowledge the tremendously important role that the people at the Center for Limnology have played? I'm afraid that I can't possibly express my gratitude in a manner that would do everyone justice. Paul Hanson has provided endless computer assistance, transformed my monte carlo programme into

Visual C++, and was a valued friend. Glen Lee not only built equipment, but most importantly provided stories, jokes, and car advice. The office staff, Linda, Alice, Sue,

Mindy, and Joe were indispensable.

With regard to students and post-docs, it's impossible to pin down individual thank yous. They have been my Wisconsin family, colleagues who critiqued my ideas, helped me put together talks, reviewed papers, shared their ideas, gave advice, provided friendship, and kept me in line when I got too out of control. In particular, I can't imagine how things would have been without Beth Sanderson, Janet Fischer, Kathy Webster, Jen Klug, Tara Reed, vi David Lewis, Chris Harvey, and Tim Johnson. There is so much more I could say and so many more I could thank..... I am indebted to all those who have made the Center for

Limnology such a wonderful place to learn and grow.

During the final year of my dissertation, I took a 9-month 'sabbatical' to teach at

Michigan Technological University in Houghton Michigan. I am indebted to John Adler, the chair of the Department of Biological Sciences, for providing me this opportunity. My year at Michigan Tech was stressful, challenging, overwhelming, and overall completely wonderful! It was an incredible learning experience. I discovered how to juggle teaching, research, and other faculty responsibilities through strategies such as time management, long hours, and locking my office door (thanks Sue Bagley). Janice Glime and Tom Snyder provided lots of advice and guidance on teaching. The staff was wonderful in helping me get settled and making my job and research run smoothly (thanks Debbie, Pat, Donna, Jeff, and

Tom). Stephen Bowen was not only a running partner, he was also an important mentor.

Thanks Stephen for advice, insights into the inner-workings of a university, friendship, and those incredible 20-mile runs along Lake Superior. Finally, there were friends who provided me with wonderful memories of my time at MTU: Scot, Doug, Jeff, Paul and most of all,

Pete.

Last, but certainly not least, there were several people who indirectly contributed to this thesis by directly contributing to my overall well-being. Jennifer Morgan faithfully met with me three times a week to lift weights, talk about life, and laugh ourselves silly. She was and continues to be a constant source of inspiration. Eva Schindler has been a soul-mate for many years. She has been a source of strength and love, helping me keep my sanity by means of many phone calls, canoe trips, and wonderful vacations. Finally, I want to thank vii my family. They have all been incredibly supportive throughout my entire graduate career. I can't begin to explain their importance in this whole process. It has something to do with unconditional support, extreme generosity, never-ending love and concern, and the fact that no matter what the challenge was they believed I could handle it. PREFACE

Over the past few decades, there has been increasing concern about biodiversity

(Wilson 1986, Schulze and Mooney 1993, Richlefs and Schulter 1993, Heywood 1995,

Humphries et at. 1995). Biodiversity is generally defined as the 'variability among living

organisms from all sources and the ecological systems of which they are a part' and can

encompass scales ranging from genes to species to populations to ecosystems (Heywood

1995). Most frequently, biodiversity is equated with species richness, i.e. the number of

species in a particular habitat or area. It is at this level, that I have focussed my studies.

Human activities have had a large impact on ecosystems worldwide, with the frequent result of loss of species. This has resulted in a number of actions, including laws and policies such

as the Endangered Species Act (Endangered Species Act 1988). In addition, there have been an increased number of regional and international biodiversity initiatives (e.g., Global

Biodiversity Strategy. Guidelines for Action to Save, Study, and Use Earth's Biotic Wealth

Sustainably and Equitably. 1992, Biodiversity in Canada: a science assessment for

Environment Canada 1994).

There has also been the development of a considerable body of theory addressing the processes that control biodiversity. MacArthur and Wilson (1967) proposed the equilibrium theory of island biogeography to explain the number of species present on islands. In their view, island communities are maintained by a dynamic balance between rates of immigration from a mainland habitat and extinction on the local island. The resulting number of species inhabiting the island at any one time is thus a function of the island's isolation (how close it was to the mainland source of colonists) and its size (related to habitat heterogeneity). Since 2 their conception, these ideas have been expanded and applied to island-like habitats such as lakes and terrestrial habitat patches (Lassen 1975, Magnuson 1976, Aho 1978, Keddy 1976,

Brown and Dinsmore 1988, Lomolino 1989).

Metapopulation theory is conceptually similar to island biogeography; extinction of populations in small island-like habitats occurs, but the regional population is maintained through dispersal of individual among sites. That is, within a single habitat, populations go extinct but are later re-colonized by immigrants from nearby populations (Gilpin and Hanski

1991 ). This view of populations suggests that community composition is not static, but is an ever-changing parade of various species.

Lakes are like islands in many ways (Magnuson 1976). They are relatively discrete in their boundaries and, like islands, vary in extent of isolation. In fact, the upper Great

Lakes region contains thousands of inland lakes and resembles a negative image of an archipelago. This view of lakes as islands has inspired the application of island­ biogeography to aquatic systems (e.g., Barbour and Brown 1974, Keddy 1976, Dodson 1992,

Magnuson eta!. 1994). Despite these and many more studies, a fundamental question still remains unresolved; to what extent does the species composition of a lake change through time in systems that are relatively undisturbed by human activities?

Historically, it has been difficult to estimate changes in species composition, mostly because of the lack of long-term databases with standardized sampling protocols. Although temporal changes in species composition of aquatic systems have been measured (e.g.

Browne 1981, Magnuson et a!. 1994) we are uncertain about actual extent of community change because sampling errors influence the results. A problem that plagues any research 3 done on sampled populations is the uncertainty associated with missing species present at low density or high spatial heterogeneity. Species that fluctuate above and below our level of detection, but remain present in the habitat, would inflate species turnover rates. In contrast, species that invade, but remain at undetectable densities would contribute to an underestimation of species turnover. These problems have been recognized for quite some time (Fisher 1943, Nilsson and Nilsson 1983), but few studies account for sampling error in species turnover estimations (e.g., Schoener 1983). There have been recent attempts to account for sampling error in species turnover rates (Magnuson et a!. 1994, Nichols eta!. in press). Species turnover rate, the percent change in composition, may be underestimated when sample periods are separated by long time intervals (Diamond and May 1977). That is, species that went extinct and then re-colonized during the time interval would not be detected and therefore would not be incorporated into the turnover rate.

I addressed several questions related to the assessment of aquatic biodiversity using long-term databases collected from study sites located in the upper Great Lakes region. I estimated apparent species turnover rates (the observed change in species composition between years) for phytolankton, zooplankton, and fishes. Because apparent turnover rates consist of both actual turnover (resulting from immigration and extinction) and sampling error (resulting from missing species), I developed a method to estimate the possible turnover attributable to sampling error. To ensure that sampling error was as low as possible, I used databases where samples were taken each year for more than a decade, and most importantly, sampling protocol remained consistent throughout the study period. High apparent turnover rates have important implications for the assessment of species richness. I compared short- 4 term (single-year) and long-term (multiple-year) estimates for each taxonomic group and found that the total species pool was approximately twice the estimated richness based on a single year's sample.

Chapter one, "Crustacean zooplankton species richness: single- and multiple-year estimates" calculates species richness of zooplankton in eight Canadian Shield lakes at different temporal and spatial scales, using three methods of estimation. I found that richness was high! y dependent on the scale and method of estimation. Richness estimates increased as the number of samples increased, including spatial, intraannual, and interannual samples.

While it is generally accepted that richness increases as number of individuals encountered increases (Fisher 1943), few studies have investigated the influence of a temporal component, particularly on the scale of years (but see Magnuson et al. 1994). Changes in species composition, both within a year and among years, results in increased richness with the duration of the database. Furthermore, the relationship between short-term and long -term richness varied among lakes, depending on species turnover rates. The result was that different environmental variables explained variation in richness rankings among lakes, depending on the scale of observation. This chapter is currently in the Canadian Journal of

Fisheries and Aquatic Sciences 55:1573-1582.

Chapter two, "Inter-annual variability and species turnover of crustacean zooplankton in Shield lakes" demonstrates that apparent species turnover rates for zooplankton are high, averaging 16% per year. Apparent turnover rates varied among years and were influenced by the number of years investigated, the interval between samples, and the occurrence of rare 5 species. Because apparent turnover consists of both sampling error and actual turnover attributable to immigration and extinction of species, I developed a method for estimating the possible contribution of sampling error to turnover rates. I found that potential sampling error was high and difficult to distinguish from apparent tumover. This chapter is currently in press in the Canadian Journal of Fisheries and Aquatic Sciences.

Chapter three, "Long-term species turnover and richness estimates: a comparison among aquatic organisms", compares turnover and richness estimates of phytoplankton, zooplankton, and fishes from several lakes in three study sites in the upper Great Lakes region. Based on life-history traits and dispersal mode, I predicted that turnover would be greatest for phytoplankton, intermediate for zooplankton, and lowest for fishes. Apparent turnover was similar among all three trophic levels; 18% for phytoplankton, 16% for zooplankton, and 20% for fishes. Sampling turnover, however, was high for all three groups.

In fact, sampling turnover was of the same magnitude as apparent turnover such that comparisons among taxonomic groups were problematic. This chapter is current! y being prepared for submission to an ecological journal.

Chapter four, "Lakes as Islands: biodiversity, invasion, and extinction" puts this work into a historical and present-day context of biodiversity research at the North Temperate

Lakes Long-Term Ecological Research site. In 1976, John Magnuson suggested that lakes were similar to islands and used the framework of island biogeography (MacArthur and

Wilson 1967) to understand community dynamics in lakes. Since then, Magnuson, Stanley

Dodson and several students, including Bill Tonn, Frank Rahel, Rigel Cisnernos, Ann 6 McLain, Tom Hrabik, and I have continued to develop these ideas by investigating factors that control immigration and extinction rates and ultimately, species richness and assemblage composition. This chapter summarizes these contributions and will be published in "Lakes in the Landscape: Long-Term ecological Research of North Temperate Lakes", edited by J. J.

Magnuson and T.K. Kratz. 7

References

Aho, J., 1978. Freshwater snail populations and the equilibrium theory of island biogeography. I A case study in southern Findland. Annales Zoologici Fennici 15:146-154.

Biodiversity in Canada: a science assessment for Environment Canada. 1994. Biodiversity Assessment Team. Environment Canada, Ottawa.

Barbour, C. D. and J. H. Brown, 1974. Fish species diversity in lakes. The American Naturalist 108:473-489.

Brown, M. and J. J. Dinsmore, 1988. Habitat islands and the equilibrium theory of island biogeography: testing some predictions. Oecologia 75:426-429.

Browne, R. A., 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. Journal of Biogeography 8:75-83.

Dodson, S. 1992. Predicting crustacean zooplankton species richness. Limnology and Oceanography 37:848-856.

Diamond, J. M. and R. M. May, 1977. Species turnover rates on islands: dependence on census interval. Science 197:266-270.

Endangered Species Act. 1988. Endangered Species Act of 1973. As Amended through the 1 100 h Congress. United States Fish and Wildlife Service, United States Department of the Interior. Washington, D.C., USA.

Fisher, R. A., A. S. Corbet and C. B. Williams, 1943. The relation between the number of species and the number of individuals in a random sample of an animal population. Journal of Animal Ecology 12:42-58.

Gilpin, M. and I. Hanski (eds.). 1991. Metapopulation dynamics: empirical and theoretical investigations. Academic Press, San Diego. 336 p.

Heywood, V. H., Executive Editor, 1995. Global Biodiversity Assessment. UNEP. Cambridge University Press, Cambridge. 1140 p.

Humphries, C. J., P. H. Williams and R.I. Vane-Wright, 1995. Measuring biodiversity value for conservation. Annual Review of Ecology and Systematics 26:93-111. 8 Keddy, P. A., 1976. Lakes as islands: the distributional ecology of two aquatic plants, Lemna minor L. and L. Trisulca L. Ecology 57:353-359.

Lassen, H. H., 1975. The diversity of freshwater snails in view of the equilibrium theory of island biogeography. Oecologia 19:1-8.

Lomolino, M. V., 1989. Interpretations and comparisons of constants in the species-area relationship: an additional caution. The American Naturalist 133:277-280.

MacArthur, R. H. and E. 0. Wilson, 1967. The Theory of Island Biogeography. Princeton University Press, Princeton.

Magnuson, J. J., 1976. Managing with exotics-a game of chance. Transactions of the American Fisheries Society 105:1-9.

Magnuson, J. J., B. J. Benson and A. S. McLain, 1994. Insights on species richness and turnover from long-term ecological research: fishes in north temperate lakes. American Zoologist 34:437-451.

Nichols, J.D., T. Boulinier, J. E. Hines, K. H. Pollock and J. R. Sauer, in press. Estimating rates of local species extinction, colonization and turnover in animal communities. Ecological Applications

Nilsson, S. G. and I. N. Nilsson, 1983. Are estimated species turnover rates on islands largely sampling errors? American Naturalist 121:595-597.

Richlefs, R. E. and D. Schluter, editors. 1993. Species Diversity in Ecological Communities: Historical and Geographical Perspectives. University of Chicago Press, Chicago. 414 p.

Schoener, T. W., 1983. Rate of species turnover decreases from lower to higher organisms: a review of the data. Oikos 41:372-377.

Schulze, E.-D. and H. A. Mooney (eds.). 1993. Biodiversity and Ecosystem Function. 99. Springer-Verlag, Berlin. 525 p.

Wilson, E. 0. (ed.). 1986. Biodiversity. National Academy Press, Washington, D.C. 521 p.

WRI/IUCN/UNEP 1992. Global Biodiversity Strategy. WRIJIUCN/UNEP, Washington, DC. 9 CHAPTER ONE

CRUSTACEAN ZOOPLANKTON SPECIES RICHNESS: SINGLE- AND MULTIPLE-

YEAR ESTIMATES 1

Abstract

Richness estimates are dependent on the spatial and temporal extent of the sampling programme and the method used to predict richness. We assessed crustacean zooplankton richness in eight Canadian Shield lakes at different temporal and spatial scales, using three methods of estimation: cumulative, asymptotic, and Chao's index. Percent species detected increased with the number of spatial, intra-annual, or inter-annual samples taken. Single samples detected only 50% of the annual species pool and 33% of the total estimated species pool. This suggests that previous estimates of zooplankton richness, based on single samples in individual lakes, are too low. Our richness estimates for individual lakes approach the total number of zooplankton found in some regions of Canada, suggesting that each lake has most taxa at some time, the majority being very rare. Single-year richness estimates provided poor predictions of multiple-year richness. The relationship between richness and environmental variables was dependent on the method of estimation and the number of samples used. We conclude that richness should be treated as an 'index' rather than an absolute and sampling efforts should be standardized. We recommend an asymptotic

1Arnott, S. E., J. J. Magnuson, and N.D. Yan. 1998. Canadian Journal of Fisheries and Aquatic Sciences 55:1573-1582. 10 approach to estimate zooplankton richness because the number of samples taken influenced it less.

Introduction

Increased biological impoverishment of freshwater ecosystems, resulting from human activities such as habitat destruction, release of toxic materials, and introduction of exotic species, has heightened the importance of assessing biodiversity and understanding its connections with ecosystem function (Carpenter et al. 1992; Schulze and Mooney 1993).

Obtaining a reliable and repeatable estimate of biodiversity is a critical element in all such endeavors. However, many issues associated with developing acceptable methods of estimating the number of species in a particular habitat or ecosystem are unresolved (Harper and Hawksworth 1994).

One such issue is that biodiversity is a property that can't be measured with absolute certainty because we can rarely census an entire community. Therefore, we must determine the sampling effort and temporal and spatial scale at which biodiversity should be assessed.

Sampling issues may be particularly important in aquatic habitats because of their small-scale patchiness and short-term temporal variability. Methodological factors certainly influence estimates of species richness in aquatic habitats (Minns 1989; Magnuson et al. 1994).

Estimated species richness increases with the number of individuals or samples collected (Fisher et al. 1943). Several techniques that exploit this relationship have been developed, including rarefaction (standardizing to a common number of individuals)

(Hurlbert 1971), species accumulation curves (Miller and Wiegert 1989), species-area curves, integration of lognormal distributions, and jackknife and bootstrap re-sampling II methods (Heltshe and Forrester 1983). These methods are primarily directed toward addressing the influence of increasing the number of individuals collected, but, few studies have explicitly investigated how richness estimates are influenced by the spatial and temporal extent of the sampling routine.

Zooplankton populations exhibit substantial spatial (e.g. Pinel-Alloul and Pont 1991) and temporal heterogeneity (Sommer eta!. 1986). The primary goal of this paper is to evaluate how this patchiness influences assessments of species richness in lakes. Because of this patchy nature of zooplankton populations, we would expect that richness estimates would be highly influenced by both the spatial and temporal extent of the sampling programme. Single samples taken at one location in the lake or on one date may not adequately represent the total species richness of the community.

We present estimates of crustacean zooplankton species richness at several different temporal and spatial scales; multiple stations within a lake, multiple samples within a year, and multiple-year samples. From the results of these comparisons, we suggest a method of estimating zooplankton species richness and develop a strong rationale for the idea that standardized methods must be used to arrive at indices. That is, sample richness and total richness do not converge, even with a relatively large number of samples and the difference between them is highly dependent on the number of samples taken. The level of sampling effort must be standardized to obtain meaningful comparisons among sites. 12 Methods

Study lakes

The eight study lakes are located on the Canadian Shield in South-central Ontario,

Canada near the Dorset Environmental Science Centre. The watersheds are forested (mixed deciduous and coniferous) and non-agricultural, owing to rough topography, thin soils, and cool climate. Lakes differ in characteristics important to the ecology of Shield lakes, including size, trophy, alkalinity, presence of glacial relicts, and invertebrate predator and fish populations (Table 1).

Zooplankton sampling

Although the sampling protocol was not designed for biodiversity research (data were collected as part of a biomonitoring programme), we believe analysis of this database can provide useful insights into sampling issues associated with the estimation of species richness. The primary strength of the Dorset zooplankton database is that methods have remained consistent since 1978, a longer period than for most studies. Lakes were sampled at the same location using the same gear, deployed in the same manner under the direction of the same crew chief. Samples were preserved, composited, and sub-sampled in the same way, and counted using the same protocol by the same person (Yan et al. 1996).

Each lake was sampled from 6 to 27 times each year during the ice-free season from

1978 to 1989. For multiple-year analyses, we standardized the database to six samples per year for each lake. When more than six samples were taken in a year we randomly chose one sample per month during the ice-free season. When there were more than six months when 13 samples were collected, we used data from months closest to the summer season (May to

October).

Zooplankton were collected with a metered plankton net (McQueen and Y an 1993) at a single station located at the deepest area in the lake. The conical net had a mesh size of 80

1-1m and a mouth diameter of 12.5 em. It was equipped with a uni-directional current meter.

Sample volumes were calculated using measured haul filtration efficiencies, which averaged

81%.

The net was hauled from several different depths to the surface (Yan et a!. 1996) and the contents of the hauls were combined. There are large, often persistent, vertical differences in zooplankton abundance, and community composition is often related to thermal stratification. Samples were composited such that vertical strata contributed to the sample in proportion to strata volumes. This was done so that samples would not be biased toward organisms occupying lower depths nor would the epilimnion be under-represented.

Samples were preserved with 6% buffered sugar-formalin. A minimum of 250 crustacean zooplankton was counted in each sample. A Folsom plankton splitter was used to generate sub-samples of various sizes. The larger sub-samples were used to count rare, large animals, whereas the smaller sub-samples were used to count the most abundant species. In this way, no single species contributed more than 20% of the total count. All Cladocera and mature

Copepoda were identified to species. We combined counts of Diaphanosoma brachyurum with D. birgei to allow for changes in nomenclature over the course of the study (Korfnek

1981). Daphnia pulicaria and D. pulex were combined and enumerated as D. pulex because we did not believe they could be reliably distinguished (Dodson 1981). Eubosmina 14 longispina was excluded from our analysis because its identification was not consistent throughout the entire database. Nauplii and other unidentifiable juveniles (mostly copepods)

were excluded from our analyses.

Species richness was estimated at several temporal scales, using three different methods of calculation. For clarity, we use a standard notation throughout the paper: S • years.

# ,.mples within a year. index· For example, S I2.6.cum is based on 12 years and six samples within each year using the cumulative richness estimate.

Single-Year Richness Multiple stations within a lake

Ten stations, including the central station, were sampled once in May, June, July, and

August, 1985 in Plastic Lake. The stations were uniformly distributed throughout the lake at

depths ranging from 5-16 m. A single integrated sample was taken by towing a plankton net

from the bottom of the lake to the surface at each station. Therefore, these samples differed

from the long-term samples in that they were from a single tow and were not volume-

weighted. Preservation and enumeration followed the methods described above. Total

annual richness (S I.4.cum) was determined for individual stations, in addition to a composite

richness for all 10 stations.

To determine the influence of increasing the number of stations, we calculated

cumulative richness for varying numbers of stations for each sampling date. We calculated

richness by randomly selecting sub-samples comprised of from I to 10 stations and

calculating a total species list for each sub-sample. This randomization procedure was

repeated 1000 times to determine a mean cumulative richness and standard deviation. 15

Multiple samples within a year

In the entire database, there were 9 occasions when a lake was sampled more than 20 times within a year. We used these data to investigate how increases in the number of intra­ annual samples influence the estimated species richness. Annual species richness (S 1.;.cum) was estimated for each lake-year by sub-sampling the database, without replacement, I to 20 times. We drew one sample from every month before a second from any particular month could be drawn to prevent temporal clumping. We calculated species richness as the total number of species in the sub-sampled database. This procedure was repeated 1000 times to obtain a mean richness estimate and standard deviation.

We used the 1980 database to compare the performance of three indices of species richness; cumulative, asymptotic richness and Chao's Index. In 1980, each lake was sampled at least 17 times. Cumulative species richness (Scum) was the total species pool for a particular number of samples. Cumulative species richness for different numbers of samples from I to 17 was obtained by repeating this analysis for all possible combinations.

The asymptotic richness (Sasym), based on a Walford plot (Ricker 1975), was obtained by calculating where mean cumulative richness (Scum) at sub-sample size t versus mean cumulative richness at sub-sample size t+ I intersects the I: I line (Fig. 1).

Chao's index provides a relatively unbiased non-parametric estimate of total species richness (Colwell and Coddington 1994) and has been used in other studies to estimate zooplankton richness (Dumont and Segers 1996). The number of species in the lake, S* was estimated based on the occurrence of rare species according to: !6

S* =Sobs+ (L2 I 2M)

where Lis the number of species that occurred in only one sample, M is the number of species that occurred in exactly two samples, and Sobs is the observed richness (Chao 1984).

We estimated Chao's richness (Schao) using sub-sample sizes ranging from 2 to 17.

Multiple-Year Richness

To estimate long-term species richness we used cumulative species richness (Scum).

Chao's index (Schao). and asymptotic richness (Sasym). Individual sub-samples, however, were based on annual richness estimates obtained from six monthly samples (S 1,6,cum). We used six samples because this was the largest number of samples consistently taken throughout the entire database.

Comparison of Richness Estimates

We compared richness ranking among lakes obtained for single year and multiple year richness estimates. Spearman rank correlations were calculated, comparing each of the three methods for estimating richness (cumulative, asymptotic, and Chao's Index), single­ year and multiple-year richness, and one, six and 12 sub-samples.

In an attempt to identify why correlation of single-year and multiple-year richness was low, we calculated the Spearman rank correlation coefficient for asymptotic (S 12,6,asym) and mean annual species richness (S 1,6,asym) in all eight lakes under several different scenarios; 1. We removed singlets from the database (species that occurred in only one 17 year), 2. Typical littoral species were removed, 3. Rare species (with a probability of detection, P < 0.25) were removed.

where P is the probability of detecting at least one individual assuming that species approximated a negative binomial distribution across time, m is the number of individuals in

2 2 the sample, and k is a parameter that is estimated as k=x /(s - x), where xis the sample mean density and s2 is the sample variance.

Gradients in richness estimates are frequently compared with environmental gradients to determine factors controlling community structure (pH: Y an et a!. 1996, habitat area:

Browne 1981; Yan eta!. 1996, productivity: Dodson 1992, and water hardness: Carteret a!.

1980). However, the relationship between environmental variables and richness may be influenced by the method and temporal scale used to estimate richness. We compared several explanatory variables for both single-year and multiple-year richness estimates. The importance of log area, maximum depth, conductivity, pH, total phosphorus, and flushing rate in accounting for variation in richness among lakes was evaluated using a forward, stepwise multiple regression. None of these variables were significantly correlated to each other, using a Bonferroni adjustment for multiple comparisons (Table 2). 18

Results

Estimation of Single-Year Richness

Multiple stations within Plastic Lake

Annual richness (S 1, 4,cum) was higher at the central station ( 16 species) than at all other stations (12 to 15 species, mean=l2.8). This occurred probably in part, because the central station is deeper and therefore, the sample represents a greater volume of water and also includes hypolimnetic habitat not occurring at some of the shallower stations. We found· that the number of species detected increased with depth sampled (P = 0.001, R2 = 0.29), although there was considerable variation in the relationship. The central station, however, under-estimated the total species pool, based on the cumulative richness of all stations, across all months (16 species at the central station versus 21 total species at all stations). Four of the five species that were not detected at the central station were detected at only one station on one date and typically occurred at low densities (<0.12/L).

Increases in the temporal extent of sampling (more samples within a year) and increases in spatial extent (more sampling stations) resulted in higher cumulative richness

(Fig. 2). Cumulative richness was similar for a given number of samples for both sampling methods, except that in August spatial samples resulted in slightly higher cumulative richness estimates. For May, June, and July cumulative species richness was the same after 9 samples had been taken. 19

Multiple samples within a single year

A single sample (Sl,l), randomly taken during the ice-free season, resulted in the detection of less than 50% of the total species pool obtained from 20 samples (S 1,2o,cum) taken throughout the ice-free season (mean 48.6% ± 9.9 S.D.). Species richness of single samples taken in 1980 ranged from 8 to 13 species in the eight·lakes (Table 3). The coefficient of variation among sample dates ranged from 14 to 26% and averaged 16%. Species richness

(S l,l), estimated from indiviifual samples, was compared among months to determine if any month provided a best estimate of annual species richness. Although there was no significant difference in the percent species detected each month (ANOV A, P=0.36), the greatest number of species was detected in June (55% of total richness based on 20 samples) and the least were detected in November (39%) (Fig. 3).

More intra-annual samples were required as the desired level of species detection increased (Fig. 4 ). The number of samples required to detect between 50 and 90 percent of species present in a year, increased rapidly fro'm 2 to 10. To detect 80% of the total species pool, at least 5 samples (±2) needed to be examined (Fig. 4). This was approximately equivalent to monthly sampling during the ice-free season. To detect 95% of the species, 13 samples (±3) were required (Fig.4). This required a biweekly sampling regime during the ice-free season.

Richness estimates derived from Chao's index were always greater than estimates obtained using asymptotic and cumulative methods (Table 3, Fig. 5). In general, cumulative and asymptotic richness estimates were similar, although cumulative richness was frequently 20 higher when 12 or more samples were taken and asymptotic richness tended to be higher

when fewer than 9 samples were taken. Variation among asymptotic richness estimates based on different numbers of samples was low (mean = 7% ). In contrast, the coefficient of variation among estimates of Chao's richness for different numbers of samples was highest

(Chao mean= 23%, cumulative mean= 16%).

Multiple-Year Richness

There was considerable temporal variation in the presence of species from year to year (Table 4). A single-year sample, based on six samples per year (S 1•6.cum) predicted 60%

of the total observed species pool for 12 years (S 12•6.cum). Although annual richness under­

estimated the long-term richness, it was consistent from year to year (mean coefficient of

variation across years: 13%).

For each richness index used, richness estimates were highly dependent on the

number of samples taken (Fig. 5). Chao's Index almost always predicted a higher richness

than the other two indices, regardless of sample size. Chao's estimates increased the most

between l and 4 years. When more than 4 years of samples were used to estimate richness,

variation in richness estimates was much smaller; the coefficient of variation among number

of years sampled was reduced from 21% when all samples were included to 9% when the

number of years was greater than 4. Asymptotic richness was always higher than cumulative

richness estimates. The coefficient of variation among richness estimates obtained from

different numbers of samples was highest for Chao's index (mean of lakes= 21 %),

intermediate for cumulative (mean of lakes= 15%), and lowest for asymptotic richness 21 (mean of lakes= 12%). Variation in richness estimates was generally highest for Heney

Lake. This results from the occurrence of a large number of episodic species that appeared in on! y one year.

Correlation among Richness Estimates

Spearman rank correlation of richness estimates of lakes among single-year sampling regimes was high. That is, estimates based on single years were similar regardless of the particular sampling regime used (Table 5). Similarly, multiple-year species richness estimates, regardless of regime, were similar to each other. In contrast, correlations between single-year and multiple-year estimates were extremely low and frequently negative. That is, the number of species found in a single-year was not a good predictor of the number of species found in multiple-years. In general, the correlations between richness rankings were highest when numbers of samples were equal, regardless of the index type (mean r, = 0.87 for comparisons between multiple-year samples and mean r, = 0.79 for comparisons between single-year samples). When comparisons were made using the same index, but different numbers of samples, correlation was low (mean r, = 0.43 between single-year samples, mean r, = 0.21 between multiple-year samples, and mean r, = -0.03 between single and multiple­ year samples). Single-year richness estimates (e.g. Su) did not correlate with long-term estimates (e.g. S,z,6.cum• S12.6.asy. or S 12,6,chal (Spearman rank correlation= -0.49, -0.55, -0.60, respectively). Correlation coefficients between S 12,6,asymand S1,6.asym improved when rare species were removed. The greatest improvement was obtained when singlets (species that occur in single years) were removed (Spearman rank correlation= 0.04 when all species • 22 were included, 0.97 when singlets were removed, -0.06 when littoral species were removed, and 0.36 when species with less than a 0.25 probability of detection were removed). When episodic species were removed from the analyses single-sample richness was a good predictor of the long-term ranked richness among lakes.

Explanatory Variables of Richness

The relationship between species richness and environmental variables was dependent on both the index type and the number of years used to estimate richness. Depth explained 61% of the variation among lakes in single-year estimates of richness (S 1,6,cum).

Lake flushing rate, log area, conductivity, total phosphorus, and pH did not contribute to the variation in richness among lakes. When other indices were used to calculate richness different results were obtained. Conductivity explained 65% of the variation in single-year asymptotic richness (S 1.6.u,ym) (P = 0.02) and none of the environmental variables explained variation in Chao's single-year richness (S1,6,chuo). Variation in multiple-year richness was explained by total phosphorus when either cumulative or asymptotic richness was estimated

2 2 (SI2,6.cum: P = 0.07, R = 0.45, S12.6,usym: P =0.095, R = 0.40). None of the environmental variables explained a significant amount of the variation in richness estimated using Chao's index (S 12,6,chao).

Discussion

Our estimates of total species richness, based on an extensive 12-year database, were considerably higher than expected from two previous multi-lake surveys. Dodson (1992) predicted that between 8 and 10 planktonic crustacean species should be found in lakes 23 within the size range of Dorset lakes. Similarly, Patalas (1990) suggested that lakes in

Northwest Ontario would have, on average, 10.7 species. We estimated the Dorset lakes to have mean annual richness from 11.4 to 17.3 species and total species pools ranging from 21 to 40 species, based on multiple-year asymptotic richness (Table 3). Our values are higher than Dodson's and Patalas' estimates for two reasons; our estimates were based on more samples, collected over an extended period (6 monthly samples* 8 to 12 years) and we didn't eliminate typically littoral species that were caught in our pelagic samples (as in

Dodson's study). The lakes for both Patalas' and Dodson's studies were sampled infrequently; single samples in Patalas' study and a minimum of three samples in Dodson's study. When single sample surveys were used to estimate species richness in Dorset area lakes, results similar to Dodson's and Patalas' estimates were obtained: Yan and Strus (1980) found 9.3 to 14.6 species per sample on average in Blue Chalk, Jerry, Chub, Harp and Dickie

Lakes in 1977. Keller and Yan (1991) reported a mean of 10 species per sample in 24, non­ acidic (pH>6.0) Dorset lakes. Therefore, with comparable sampling effort, our richness estimates for the Dorset lakes are similar to previous studies. However, our richness estimates based on long-term data suggest that it is likely all of these investigators underestimated the actual size of the species pool. It is interesting to note that our richness estimates based on many samples in a single lake, approach Patalas' estimates for the number of taxa in many lakes in a region of Canada (Patalas 1990). One possible explanation is that most species within a region may be present in each lake at some point in time, although they are very rare. 24 Our comparison of spatial and seasonal sampling in Plastic Lake suggests that increasing the temporal or spatial extent of sampling improves the richness estimate. The identity of the species, however, varied depending on sampling regime. Eubosmina tubicen,

Holopedium gibberum, Leptodiaptomus minutus, and Daphnia ambigua were detected only in the temporal samples; and Alana sp., Chydorus sphaericus, Daphnia retrocurva, and

Eurycercus lamellatus were collected only in the spatial samples. Therefore, the decision to choose one sampling regime over another depends on the goals of the study. Although species richness estimates were similar, multi-station samples tended to detect more littoral species, whereas the multi-date samples tended to detect typically pelagic species with temporally variable abundances. This result is probably dependent on lake size. Patalas and

Saiki (1993) found that the proportion of total species detected at a single sampling station was dependent on lake size, with a higher proportion of the total species being detected in smaller lakes. Similarly, fewer spatial samples were required in small lakes ( <170 ha) to capture 90% of the species, although Lake 373 (27 ha) was a notable exception to this trend.

Therefore, the value of spatial versus temporal sampling may depend on the size of the lake.

All of the Dorset lakes used in our study are small, ranging from 20 to 100 ha in size. Had we used a larger lake than Plastic Lake, we may have found a higher importance of spatial samples relative to temporal samples.

Mean annual richness did not predict richness based on multiple-year samples. This result was not surprising as Hurlbert (1971) suggested it as a general possibility in his review of species diversity measures. In our study lakes, as the temporal extent of sampling increased, total richness also increased, probably because of increased sampling effort (i.e. 25 identifying a larger number of individuals) and because of temporal oscillations in species abundances. Zooplankton undergo both seasonal shifts in population abundances (Sommer eta!. 1986) and inter-annual changes in species composition (Browne 1981). Single samples or single-year samples failed to detect highly variable species that were present infrequently throughout the entire sampling record. Correlations between single-year and multiple-year richness were low and almost always negative. Some lakes with low species richness in single-year estimates had high multiple-year richness. For example, Heney Lake was tied for the fewest species when richness was assessed on an annual basis. When richness was determined by extrapolation from long-term data, however, Heney Lake had the most species. Heney Lake is unique in that it has a high species turnover rates; approximately

25% of the species in Heney Lake changed from year to year (S. Arnott, unpublished data).

Mean annual richness remained relatively constant but the total species pool increased annually. In contrast to this, however, other lakes maintained their relative ranking for both short-term and long-term richness estimates. Blue Chalk Lake had both a low short-term richness (S u) and a low richness based on long-term estimates (S9,6,cum). This discrepancy arises, in part, because in different lakes species may be appearing and disappearing from year to year at different rates. The difference between mean annual species richness (S 1•6.cum) and multiple-year richness (S l2,6,cum) was highly correlated with measured species turnover

(S. Arnott, unpublished data) (linear correlation: r=0.868, P=0.005).

Much of the difference between short-term and long-term richness arises, in part, from species appearing for single years. When singlets were removed, the correlation between mean annual and long-term richness improved. Most of the singlet species occurred 26 at relatively low densities (mean= 0.014 individuals/litre± 0.018 S.D.) compared to the overall mean density of species (mean= 0.489 individuals/litre± 1.187 S.D.) and it is uncertain what role they play in the community. They may be opportunistic species that occasionally arise when certain environmental conditions are met, or they may be species that are always present but are undetectable using typical sampling and counting regimes.

These rare species may function as 'ecological memory', enabling the community to rapidly respond to environmental changes (Padisak 1992). Therefore, the decision to use single-year richness versus long-term richness should be at least partially based on the importance of rare, episodic species to the hypotheses being tested. If rare species are removed from the analyses there is good correlation between single-year richness and multiple-year richness, indicating that single-year estimates may provide useful estimates of the species pool.

As a result of these temporally-dependent switches in richness rankings, there were differences in how short-term richness and long-term richness correlated with environmental variables. There were also differences in correlations, depending on indices used for richness estimates. Depth was an important explanatory variable for cumulative annual richness

(S!,6,cuml and conductivity explained some of the variance in asymptotic annual richness

(S!,6,asym), For long-term richness estimates (S12,6,cum and S 12,6,asym) total phosphorus was the

only significant explanatory variable. In addition to the influence of study duration and

richness measure, these results are also dependent on the number of lakes in the analysis.

Using a larger set of lakes (47) located in the same region as our eight study lakes, Yan eta!.

(1996) found that single-year richness estimates were correlated with lake acidity and size.

Despite our small sample size (8 lakes), the discrepancy among richness estimates derived 27 from a different method suggests that care must be taken when drawing conclusions from

species richness data. Magnuson et al. (1994) drew a similar conclusion based on analyses of

fish species richness in 7 Wisconsin lakes. The coefficient of the species-area relationship was dependent on whether mean annual richness or cumulative richness was used in their paper. As with Magnuson et al. ( 1994 ), we conclude that richness determined by sampling

should be treated as an 'index' rather than total richness. Thus, care must be taken to use

standardized sampling efforts and estimators, especially across time because duration of the

study influences our perception of the relationship between species richness and explanatory

environmental variables.

Annual richness and multiple-year richness estimates provide different information

regarding the zooplankton community. Generally, richness estimates based on a single

sample represent the number of potentially interacting species in a lake or the number of co­

occurring species. It is possible that some of the species may have little or no effect on each

other, but there is potential for direct or indirect interactions. It is also possible that species

that don't overlap in time may also influence each other through events such as delayed

changes in resource quality or supply. However, as a first cut, annual richness may be good

representation of the number of species with the potential for interaction. Annual richness

measure may be useful when quantifying the effect of a change in chemistry or the influence

of an invading species on community structure. Because the probability of capture is related

to abundance, rare species will probably be overlooked. However, it may be justifiable to

assume that rare species play a minor role in food web interactions at a given moment in

time. Mean annual richness based on several samples throughout the season increases the 28 probability of detecting rare species and also facilitates the detection of seasonally dynamic

species. Multiple-year richness represents a longer time scale and takes into account shifts in

species composition. It is a measure of the potential species pool. This type of information

may be useful in biodiversity studies where researchers are investigating the relationship between richness and stability or resilience. Knowing the total species pool could provide an

indication of the potential for functional complimentarity (sensu Frost et a!. 1995) within the

community. Two habitats with the same mean annual species pool may have very different

long-term richness estimates, which could lead to differences in stability. For example, in

both Blue Chalk and Chub Lakes we detected approximately 15 species each year (S 1,6.curn).

The asymptotic richness (long-term richness) was much greater for Chub Lake than for Blue

Chalk Lake (40 versus 24 species). Because Chub Lake has a large species pool, we might

expect that it would be more resistant to disturbance through functional compensation of

species (e.g. Frost eta!. 1995). That is, species that are negatively influenced by a stress may

be replaced by more resistant species with complimentary community functions.

Recommendations

There are many factors that influence the decision to choose one sampling and

counting routine over another. It will depend, not only on funding resources available, but

also the objectives of the study. For some questions, assessing the number of interacting

species at a given point in time is of primary concern. For example, single-sample or single­

year estimates may be adequate for determining the relationship between diversity and

primary productivity or predator density. Presumably, these factors would be important 29 drivers determining the number of interacting species and therefore their influence should be

detected at a temporally restricted scale. In contrast, multiple-year richness estimates would

be more important in determining differences among lakes in their potential response to

stress. Lakes with large species pools, either as rare 'memory' species (sensu Padisak) or as temporarily dormant resting stages, may be more resistant to perturbations. In this case, it would be important to extend the sampling schedule for as long as possible. However, in either situation there is the underlying importance of standardized sampling regimes.

Because richness estimates were influenced both by the number of intra-annual and inter­

annual samples, methodology should, at least, be consistent among lakes within a survey.

Differences in the method used to predict richness are also worth consideration. All three indices, cumulative, asymptotic, and Chao's were dependent on the number of samples taken. However, the asymptotic method of estimating richness provided both an estimate of the potential species pool and provided consistent values regardless of the number of years

sampled. Therefore, we recommend use of the asymptotic method of richness estimation for

multiple-sample studies.

Acknowledgments

We thank Trevor Pawson, Robert Girard, Martyn Futter, and Paul Hanson for technical assistance and Bill Geiling for counting the zooplankton samples. Tim Essington,

Kathy Webster, Tom Frost, Kazimierz Patalas, and Ben Bailey provided helpful reviews.

Craig Stow and Conrad Lamond provided advise on analyses. The Dorothy Powers Grant

and Eugene Lodewick Grant Scholarship Fund provided financial support for S.E.A ..

Support for research was provided by the Ontario Ministry of the Environment and Energy 30 and the North Temperate Lakes Long-Term Ecological Research project funded by the

National Science Foundation, Grants BSR14330 and DEB9011660.

References

Browne, R. A. 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. J. Biogeogr. 8: 75-83.

Carpenter, S. R., Fisher, S. G., Grimm, N. B., and Kitchell, J. F. 1992. Global change and freshwater ecosystems. Annual Review of Ecology and Systematics. 23:119-139.

Carter, J. C. H., Dadswell, M. J., Roff, J. C., and Sprules, W. G. 1980. Distribution and zoogeography of planktonic crustaceans and dipterans in glaciated eastern North America. Canadian Journal of Zoology. 58:1355-1387.

Chao, A. 1984. Non-parametric estimation of the number of classes in a population. Scand. J. Stat. 11: 265-270.

Colwell, R. K. and Coddington, J. A. 1994. Estimating terrestrial biodiversity through extrapolation. Phil. Trans. R. Soc. Land. B. 345: 101-118.

Dodson, S. I. 1981. Morphological variation of Daphnia pulex Leydig (Crustacea: Cladocera) and related species from North America. Hydrobiol. 83: 101-114.

Dodson, S. 1992. Predicting crustacean zooplankton species richness. Limnol. Oceanogr. 37: 848-856.

Dumont, H. J. and Segers, H.. 1996. Estimating lacustrine zooplankton species richness and complementarity. Hydrobiol. 341:125-132.

Fisher, R. A., Corbet, A. S., and Williams, C. B .. 1943. The relation between the number of species and the number of individuals in a random sample of an animal population. J. Animal Ecol. 12:42-58.

Frost, T. M., Carpenter, S. R., Ives, A. R. and Kratz, T. K. 1995. Species compensation and complementarity in ecosystem function. In Linking Species and Ecosystems. Edited by C. G. Jones and J. H. Lawton. Chapman and Hall. pp. 224-239.

Harper, J. L. and Hawksworth, D. L. 1994. Biodiversity: measurement and estimation. Phil. Trans. R. Soc. Land. B. 345: 5-12. 31 Heltshe, J. F. and Forrester, N. E. 1983. Estimating species richness using the jackknife procedure. Biom. 39:1-12.

Hurlbert, S. H. 1971. The nonconcept of species diversity: a critique and alternative parameters. Ecology 52:577-586.

Keller, W. and Yan, N. D.. 1991. Recovery of crustacean zooplankton species richness in Sudbury area lakes following water quality improvements. Can. J. Fish. Aquat. Sci. 48: 1635-1644.

Korinek, V. 1981. Diaphanosoma birgei n.sp. (Crustacea, Cladocera). A new species from North America and its widely distributed subspecies Diaphanosoma birgei spp. lacustris n. ssp. Can. J. Zoo!. 59: 1115-1121.

Magnuson, J. J., Benson, B. J., and McLain, A. S. 1994. Insights on species richness and turnover from long-term ecological research: fishes in north temperate lakes. Am. Zoo!. 34: 437-451.

McQueen, D. J. and Yan, N.D. 1993. Metering filtration efficiency of fresh water zooplankton hauls: reminders from the past. J. Plank. Res. 15: 57-65.

Miller, R.I. And Wiegert, R. G. 1989. Documenting completeness, species-area relations, and the species-abundance distribution of a regional flora. Ecology 70:16-22.

Minns, C. K. 1989. Factors affecting fish species richness in Ontario lakes. Trans. Am. Fish. Soc. 118: 533-545.

Padisak, J. 1992. Seasonal succession of phytoplankton in a large shallow lake (Balaton, Hungary) - a dynamic approach to ecological memory, its possible role and mechanisms. J. Ecol. 80:217-230.

Patalas, K. 1990. Diversity of the zooplankton communities in Canadian lakes as a function of climate. Verh. InternaL Verein. Limnol. 24: 360-368.

Patalas, K. and Saiki, A. 1993. Spatial variation of crustacean plankton in lakes of different size. Can. J. Fish. Aquat. Sci. 50:2626-2640.

Pinel-Alloul, B. and Pont, D. 1991. Spatial distribution patterns in freshwater macrozooplankton: variation with scale. Can. J. Zoo!. 69:1557-1570

Ricker, W. E. 1975. Computation and interpretation of biological statistics of fish populations. Department of the Environment Fisheries and Marine Service. Bulletin 191. 32 Schulze, E.-D. and Mooney, H. A. (Editors). 1993. Biodiversity and Ecosystem Function. Springer-Verlag, Berlin.

Sommer, U., Gliwicz, Z. M., Lampert, W., and Duncan, A. 1986. The PEG-model of seasonal succession of planktonic events in fresh waters. Arch. Hydrobiol. 106:433-471.

Yan, N.D. and Strus, R. 1980. Crustacean zooplankton communities of acidic, metal­ contaminated lakes near Sudbury, Ontario. Can. J. Fish. Aquat. Sci. 37: 2282-2293

Yan, N.D., Keller, W., Somers, K. M., Pawson, T. W., and Girard, R. E .. 1996. Recovery of crustacean zooplankton communities from acid and metal contamination: comparing manipulated and reference lakes. Can. J. Fish. Aquat. Sci. 53:1301-1327. Table 1: Morphometric and chemical data for the study lakes during the period 1978 to 1989. Values represent the mean of all samples taken during the ice-free season (generally biweekly to monthly) over the twelve-year study period. Standard deviations for chemical data are in parentheses.

Lake Latitude Longitude Area Mean Max. Alkalinity Conductivity DOC pH TP

(N) (W) (ha) Depth(m) Depth(m) rng!L CaC03 ([!S) (mg!L) ([!giL)

Blue Chalk 45° II' 78" 56' 52.4 8.5 23.0 4.09(0.34) 28.92 (0.67) 1.79 (0.20) 6.65 (0.15) 6.30 (0.87)

0 0 Chub 45 13' 78 59' 34.4 8.9 27.0 0.81 (0.23) 27.87 (0.65) 4.67 (0.28) 5.62 (0.14) 10.39 (1.42)

0 Crosson 45 05' 78 " 02' 56.7 9.2 25.0 0.55 (0.18) 25.99 (0.46) 4.03 (0.23) 5.58 (0.11) 11.00 (1.42)

0 0 Dickie 45 09' 79 05' 93.6 5.0 12.0 1.15 (0.41) 28.29 (0.77) 4.97 (0.33) 5.86(0.17) 11.38( 2.07)

0 0 Harp 45 23' 78 08' 71.4 13.3 37.5 3.29 (0.20) 34.40 (0.65) 3.81 (0.24) 6.29 (0.10) 7.52 (1.08)

0 0 Heney 45 08' 79 06' 21.4 3.3 5.8 0.51 (0.22) 25.03 (0.81) 2.82 (0.26) 5.81 (0.17) 7.04 (1.38)

0 0 Plastic 45 11' 78 50' 32.1 7.9 16.3 0.52 (0.19) 22.48 (0.57) 2.22 (0.14) 5.74 (0.13) 5.96 (0.98)

0 Red Chalk 45° 11' 78 56' 32.1 16.7 38.0 3.31 (0.24) 29.39 (0.40) 2.45 (0.19) 6.33 (0.16) 5.06 (0.71)

w w Table 2: Pearson correlation coefficients for explanatory variables of richness. None of the variables were significantly correlated to each other, using a Bonferroni adjustment for multiple comparisons.

log area max depth conductivity pH TP flushing rate log area 1 max depth 0.34 1 conductivity 0.60 0.70 1 pH 0.28 0.43 0.60 1 TP 0.43 -0.21 0.00 -0.61 1 flushing rate 0.10 0.20 0.14 0.71 -0.56 1

..,.w 35

Table 3: Mean species richness for each of the Dorset Lakes, estimated at several temporal scales. Three methods of estimation, cumulative, asymptotic, and Chao's index, were used. Single-year richness was calculated using 1980 data.

Single-year Multiple-year

No. of samples 3 6 9 12 15 I 3 6 9 12 cumulative Blue Chalk 9.6 14.9 18.0 19.2 20.5 21.1 15.3 18.2 20.4 22.8 23.0 Chub 9.0 14.0 17.1 19.0 20.0 21.0 15.0 18.1 20.4 22.8 26.0 Crosson 10.1 13.1 14.6 15.4 16.6 17.5 15.3 18.9 21.6 22.8 24.0 Dickie 9.5 13.4 15.6 16.7 17.7 18.5 16.3 20.8 24.4 27.3 31.0 Harp 11.0 14.9 17.3 19.4 20.8 22.1 16.0 18.4 21.1 22.8 25.0 Heney 8.1 10.8 13.1 13.9 14.6 15.2 11.4 16.6 22.8 26.0 Plastic 8.1 11.4 12.9 13.5 14. I 14.7 12.3 15.4 18.0 19.5 Red Chalk 12.3 17.8 19.6 20.3 21.1 21.8 17.3 20.2 22.3 23.5 24.0 mean 9.7 13.8 16.0 17.2 18.2 19.0 14.8 18.3 21.4 23.4 25.5 asymptotic Blue Chalk 16.4 18.6 19.4 20.3 20.8 22.7 22.2 26.1 24.1 Chub 15.8 18.1 19.4 20.1 20.8 19.4 21.6 25.1 40.5 Crosson 13.7 14.7 15.4 16.3 17.3 32.7 23.7 23.7 24.4 Dickie 14.2 16.1 16.7 17.5 18.2 23.4 26.9 29.9 38.2 Harp 18.4 18.6 20.7 21.8 23.0 19.5 25.6 25.5 30.5 Heney 11.9 15.0 14.7 15.0 15.3 27.7 42.9 32.7 Plastic 12.3 13.1 13.5 13.8 14.3 18.3 20.6 21.1 Red Chalk 19.0 19.8 20.1 20.5 21.0 22.8 23.6 24.3 24.4 mean 15.2 16.7 17.5 18.2 18.8 23.3 25.9 26.0 30.3

Chao's index Blue Chalk 20.1 21.9 26.9 3l.l 29.3 26.0 26.0 32.0 25.3 Chub 17.3 21.4 22.8 29.3 28.0 22.5 24.5 32.9 44.0 Crosson 16.2 19.5 17.9 20.3 18.3 26.7 27.5 24.5 26.3 Dickie 17.7 22.6 21.3 24.4 32.9 24.5 30.3 33.6 34.6 Harp 17.7 19.9 22.7 33.2 35.3 21.2 26.7 28.3 27.7 Heney 10.7 13.7 18.6 22.3 20.2 34.0 45.6 46.2 Plastic 17.7 19.9 22.7 33.2 35.3 20.1 22.0 22.0 Red Chalk 21.2 24.0 23.6 24.7 24.5 23.0 25.1 26.6 26.0 mean 17.3 20.4 22.1 27.3 28.0 24.8 28.5 30.8 30.6 36 Table 4: Species list of crustacean zooplankton from Dorset study lakes. Numbers indicate the percentage of years species were present in each of the eight lakes. Species name Blue Chub Crosson Dickie Harp Heney Plastic Red Chalk Chalk Acroperus harpae 8 11 Alana guttata 8 8 8 17 22 Bosmina longirostris 100 92 100 100 100 100 100 100 Ceriodaphnia lacustris 8 8 22 20 8 Chydorus bicornutus 8 Chydorus sphaericus 50 17 58 33 100 44 10 17 Daphnia ambigua 8 92 33 83 17 90 25 Daphnia catawba 33 50 67 8 11 10 25 Daphnia dubia 100 8 17 25 8 92 Daphnia galeata mendotae 100 67 100 67 100 44 20 100 Daphnia longiremis 100 100 17 8 11 40 100 Daphnia pulex 92 8 8 25 40 92 Daphnia retrocurva 25 100 33 100 92 100 42 Eubosmina coregoni 17 25 17 11 Eubosmina tubicen 42 75 75 100 100 100 100 100 Holopedium gibberum 100 100 100 100 100 89 100 100 Leptodora kindtii 33 33 33 50 25 42 Ophryoxus gracilis 8 Polyphemus pediculus 25 8 33 II 80 25 Sida crystallina 17 8 8 33 100 Chydorus globosus 11 Diaphanosoma spp. 100 100 100 100 100 100 100 100 Lepta diaptomus ashlandi 11 Lepta diaptomus minutus 100 100 100 100 100 100 100 100 Skisto diaptomus 100 8 100 17 33 10 100 oregonensis Lepta diaptomus sicilis 8 92 Epischura lacustris 100 92 83 75 92 44 30 50 Senecella calanoides 8 100 11 Cyclops bicuspidatus 100 42 100 75 100 22 50 100 thomasi Cyclops scutifer 100 100 100 100 17 11 100 100 Cyclops vernalis 17 25 25 50 58 Eucyclops agilis 17 11 Eucyclops speratus 8 8 Mesocyclops edax 100 100 100 100 92 100 100 100 Orthocyclops modestus 42 42 17 11 40 Tropocyclops prasinus 92 100 100 100 100 100 90 100 mexicanus Table 5: Spearman rank correlation matrix of single-year and multiple-year richness estimates. Cumulative, asymptotic, and Chao's estimates are presented. For the single-year samples, estimates are based on annual estimates using I, 6 or 12 samples within each year. For multiple-year estimates, estimates are based on 6 seasonal samples and I, 6 or 12 annual samples. 'Cu' represents cumulative richness, 'As' represents asymptotic richness, and 'Ch' represents Chao's estimate of richness. The first number after the prefix Cu, As, or Ch indicates the number of years on which the richness estimate is based. The second number indicates the number of samples within a year. Single-year samples Multiple-year samples Cumulative Asymptotic Chao's Cumulative Asymptotic Chao's #sam les I 6 12 6 12 I 6 12 I 6 12 6 12 I 6 12 Cu 1,1 I 0.49 0.60 0.55 0.54 0.99 0.09 0.09 0.64 0.32 -0.49 Cu 1,6 I 0.94 0.99 0.83 0.41 0.60 0.60 0.38 -0.20 -0.46 Cu 1,12 1 0.99 0.94 0.55 0.49 0.66 0.46 -0.12 -0.29 As 1,6 1 0.90 0.49 0.55 0.64 As 1,12 1 0.49 0.26 0.83 Ch 1,1 I 0.12 0 Ch 1,6 1 -0.03 Ch 1,12 I Cu 1,6 1 0.81 0.07 0.61 -0.16 1.00 0.32 -0.26 Cu6,6 1 0.35 0.75 0.09 0.81 0.61 0.06 Cu 12,6 1 0.41 0.94 0.07 0.23 0.93 As6,6 1 0.12 0.61 0.89 0.14 As 12,6 1 -0.16 -0.06 0.99 Ch 1,6 I 0.32 -0.26 Ch6,6 I 0.03 Ch 12,6 I

w -...J 38 Figure Captions

Figure 1: Example of richness estimate using a Walford Plot. Data are from Red Chalk

Lake. A line is fit through a plot of the cumulative richness at time=t versus time=t+ 1.

Asymptotic richness is determined by estimating the richness at which the fitted line

crosses the 1: 1 line.

Figure 2: Mean cumulative richness for two sampling regimes in Plastic Lake. Spatial

samples (squares) were taken in May, June, July, and August at 9 to 10 locations in

Plastic Lake in 1985. Intra-annual samples (diamonds) were taken during the ice-free

season in 1985. Both richness estimates were obtained by repeatedly sub-sampling the

database for different combinations of the same numbers of samples.

Figure 3: Mean monthly richness of Dorset Lakes, based on single samples taken in 1980.

Vertical bars represent the standard error of the mean. S1,6 indicates the mean single-year

richness, based on six samples per year.

Figure 4: The number of samples required per year to detect various percentages of the total

richness, based on 20 samples per year CS1.2o.cuml· Values represent averages based on 9

lake-year combinations. Vertical bars represent the standard error of the mean.

Figure 5: The relationship between the number of samples taken and the single-year richness

(top panel) and the multiple-year richness (bottom panel) for Blue Chalk Lake. Single- 39 year richness was based on samples taken in 1980. Richness was estimated using three different methods; Chao's index (diamonds), cumulative (squares), and asymptotic richness (triangles). 40

30 -r------,

.,.._ .;!: 25 +-'ro C/) ~ C/) / (].) / _§ 20 / () / a: / // 1:1 line 15 -jL------.-----..-----l 15 20 25 30

Richness at t

Asymptotic richness= y-intercept/(1-slope)

Fig. 1 May June en en 12 , 12 G.) .E 10 ~ T ~I 10 (.) ·;:: - 8 G.) +time > • space ;:; : i ft 6 ca :I I 4 E 4 :I (.) July August l:l 12 , 12 G.) "- - .E 10 ~ T j,.....1o £ :t " 10 (.) ·;:: 8 ~ T/l 8 G.) > ;:; 6 -1- 6 ca ~ :I 4 4 E :I (.) 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 number of samples number of samples Fig. 2

"'" 42

18 tn 16

Q)tn 14 s::: ..s::: 12 0 0: 10 8 6 ~~---+--4---~-+--~--+-~ May June July Aug Sept Oct Nov 81 ,6

Fig. 3 43

0 0 T"""

0 "0 (j) Q) (.) -Q) c-Q) 0 Ill ro .~ (.) Q) c. (/) 'i" bb !: p:; 0 Q) f'... -(.) ...Q) a.

0 (!)

~--~--~~--~~440 l() 0 l() 0 0 C\.1 T""" T""" satdwes JO JaqwnN 44

tJ> 25 tJ> C1) s:: 13 20 ·-a: C'tl ~ 15 s:: <(

10 ~--+----+--t-+-+----+---t--1 1 3 5 7 9 11 1315 17 Number of Samples tJ> 30 tJ> C1) s:: .s:: 25 -~a: E :I..

~I 20 0) s:: 0 -I 15 ~+--t--+-+-+--+---+----+-+-+---j 1 3 5 7 9 11 Number of Years Fig.5 45 CHAPTER TWO

INTER-ANNUAL VARIABILITY AND SPECIES TURNOVER OF CRUSTACEAN

ZOOPLANKTON IN SHIELD LAKES2

Abstract

We estimated apparent species turnover rates and richness of the zooplankton annually over a 12-year period in 8 lakes in south-central Ontario. Although species richness varied little among years (CV = 13%), apparent species turnover rates averaged 16% per year. This apparent turnover varied among years, and was influenced by census interval, the number of censuses, the occurrence of rare species, and by lake pH. However, Monte Carlo simulations indicated that turnover attributable to sampling error was high. That is, despite high apparent turnover rates, we cannot be certain whether interannual changes in community composition result from immigration and extinction of species because sampling error could large! y account for all apparent turnover. Regardless of the source of apparent turnover (sampling or immigrations and extinctions), high turnover rates imply that zooplankton biodiversity can be under-estimated in short-term studies because we detect a different assemblage of species every year. Only 113 of the total species pool for each lake was detected every year. Annual data under-estimated long-term species pools by 33-50%.

2 Arnott, S. E., N. D. Yan, J. J. Magnuson, T. M. Frost. 1998. The Canadian Journal of Fisheries and Aquatic Sciences. In press. 46 Introduction

A growing awareness of the link between environmental degradation and biotic impoverishment (Dobson et a!. 1997) has accelerated the pace of development of methods to assess biodiversity (Harper and Hawksworth 1994; Heywood 1995). Theoretical studies suggest that the species composition of island-like habitats, including lakes, is dynamic

(MacArthur and Wilson 1967). Nonetheless there have been few assessments of the rates of influx or disappearance of species from individual lakes. Such annual variation in species composition, if significant, could seriously affect our assessments of aquatic biodiversity.

The turnover of species over time significantly influences community dynamics in terrestrial ecosystems (e.g., Schoener and Spiller 1987). In aquatic ecosystems, it has rarely been quantified (but see McCormick and Cairns 1990; Magnuson eta!. 1994; Hecnar and

McCloskey 1996). Zooplankton turnover rates have only been evaluated once. Browne

(1981) estimated an average annual turnover of zooplankton species of 0.28%/yr over a 50 year period in several New York lakes. Because species turnover is influenced by the length of the inter-census interval (Magnuson eta!. 1994), Browne's estimates may not reflect annual changes in species composition. Species may have appeared and disappeared many times in the 50 years separating his two censuses (Diamond and May 1977).

Apparent species turnover between two censuses consists of two components: actual turnover caused by immigration or extinction of species in the system during the interval between censuses, and sampling error in either census. Sampling errors are attributable to the failure to detect a species in one or both censuses, a consequence of its rarity or 47 patchiness in either census. Such errors can either inflate or depress immigration or extinction estimates, thereby altering estimates of apparent turnover.

Actual turnover may be influenced by many demographic and environmental factors.

Demographic factors include the abundance, lifespan, body size, and other life history characteristics of individual species. Taxa with low abundance are more likely to go extinct than abundant species because birth and death rates vary (e.g., Schoener and Spiller 1987).

Species with longer lifespans (e.g., copepods vs. cladocerans) often have a lower probability of extinction; hence, turnover rates are negatively correlated with mean longevity of species

(Schoener 1983 ). Body size may be an important correlate of turnover because of its relationship with longevity. Finally, life history strategies that include resistant resting stages may also influence turnover rates. Changes in community composition may result from species that produce immobile resting stages that remain inactive for more than a year

(Hairston 1996).

Species turnover rates may also be altered by both natural and anthropogenically­ induced environmental change. Global warming is expected to alter thermal and oxygen profiles in lakes (De Stasio et al. 1996). Such physical changes can transform community composition by eliminating required habitats and influencing seasonal phenologies (Chen and Fo1t 1996). Changes in UV-B penetration (Williamson eta!. 1996), acidity (Keller and

Pitblado 1984), and trophic status (Malley eta!. 1988) of lakes all produce changes in community composition. In addition, there are many other factors, such as annual variation in weather patterns, timing of spring overturn, and summer mixing depth whose influence on species turnover has not been quantified. 48 Here we report on annual species turnover rates for crustacean zooplankton communities in 8 lakes that have been sampled over a 12-year period. These unmanipulated lakes often have been employed as reference systems for whole-lake experiments (eg. Keller eta!. 1992; Yan eta!. 1996). Most studies ignore the sampling component and assume apparent turnover is equal to actual turnover (but see Magnuson eta!. 1994, Nichols eta!. in press). We devised a means to quantify the potential contribution of sampling error to apparent turnover measurements. We have four specific objectives. We calculate apparent annual turnover rates, develop a method to estimate the contribution of error to apparent turnover rates, and demonstrate that apparent turnover estimates are related to sampling error.

Finally, we explore possible mechanisms driving turnover rates in the lakes.

Methods

Study lakes

The eight study lakes (Table 1) are located on the Canadian Shield in south-central

Ontario, Canada. Their watersheds are forested (mixed deciduous and coniferous) and have thin soils. In the past, clear-cut logging was an important industry, but the last major forest clearing occurred 60-100 years ago. Today, recreation is the primary industry. The lakes differ in size, trophy, alkalinity, presence of glacial relict species (Yan and Pawson 1997), and populations of fish and invertebrates (Table 1). They all may be characterized as having soft, slightly acidic, and nutrient-poor waters. The shorelines support between 0 (Plastic

Lake) and 125 (Dickie Lake) cottages or homes; hence, the proportion of the supply of phosphorus originating in human development ranges from 0 to 60% (Dillon and Molot '

49 1996). Six of the lakes are dimictic and the two shallowest (Heney and Dickie Lakes) are weakly stratified or unstratified. Seven are headwater lakes but Red Chalk Lake is downstream from Blue Chalk Lake. The zooplankton of the lakes has been monitored since

1978, primarily for use a reference sites in acidification studies (Y aneta!. 1996).

Although these lakes are relatively undisturbed, there have been some changes in water chemistry. Plastic Lake has acidified somewhat, the pH falling from 5.8 to 5.6 during the study (Dillon eta!. 1987). Dickie, Chub, and Heney Lakes are more acidic now than 100 years ago, based on inferences from diatom sedimentation (Hall and Smol 1996), but their pH did not decline during the study period.

Although this database was not designed for the assessment of biodiversity, it is one of the most extensive and methodologically consistent databases available in North America.

Samples were taken at least once per month each year incorporating any variation attributable to seasonal changes in species composition. Sampling and counting protocols have remained consistent since 1978. Lakes were sampled at the same location using the same gear, deployed in the same manner undyr the direction of the same crew chief (Robert

Girard from the Ontario Ministry of the Environment). Samples were preserved, composited, sub-sampled, and counted using a consistent protocol by the same person (Dr. W. Geiling of

Limnoservices Inc., Lansdowne, Ontario, see Yan eta!. 1996).

Zooplankton sampling

Lakes were sampled during the ice-free season from 1978 to 1989. The database has

12 years of continuous zooplankton records for six lakes, and 9 or 10 years for the other two 50 lakes. Zooplankton were generally sampled on a monthly basis after 1980 and more frequently in earlier years. We standardized the database to six samples per year for each lake by randomly choosing one sample from each month from May to October. Zooplankton were collected with a metered DRC (Dorset Research Centre) plankton net at a single station located at the deepest area in the lake. The conical net had a mesh size of 80 !Jill and a mouth diameter of 12.5 em. Sample volumes were calculated using measured haul filtration efficiencies, which averaged 81%.

Zooplankton were collected using a volume-weighted strategy. The net was hauled

from several different depths to the surface (Y an et a!. 1996) and the contents of the hauls . were combined. Depths were chosen such that lake strata contributed to the composite

sample in proportion to their volumes. Samples were preserved in the field with 6% buffered

sugar-formalin. A minimum of 250 animals were enumerated in each sample.

Spurred by advances in biochemical and genetic analysis, cladoceran taxonomy is

currently in a state of flux (eg. DeMelo and Hebert 1994, Taylor and Hebert 1994; Colbourne

and Hebert 1996, Hebert and Finston 1997a & b). It is not yet clear when this pace of

nomenclatural change will stabilize, nor are all the newest taxonomies readily transferable to

the examination of thousands of animals in long-term studies. In consequence, we adopted a

conservative approach to animal identification, admittedly predating the majority of these

recent genetic-based taxonomies. All Cladocera and mature Copepoda were identified to

species. Daphnia pulicaria and D. pulex were combined and enumerated as D. pulex

because we did not believe they could be reliably distinguished (Dodson 1981). Eubosmina

longispina was excluded from our analysis because its identification was not consistent 51 throughout the entire database. Nauplii and other unidentifiable juveniles (mostly copepods) were excluded from our analyses.

Each annual summary of the 6 samples was based on the examination of an average of I ,630 (S.D.±65) individuals, approximately half of which were identified to species (711 individuals, S.D. ±72). Those not identified to species were almost exclusively early stage copepodites and nauplii. Comparisons with weekly sampling indicated that our monthly protocol detected 80% of the total species present in that particular year (Arnott et a!. 1998).

Apparent turnover

Apparent turnover of crustacean zooplankton (between any two years) was estimated for each lake as

T = 100 * (I+ E)/[(St + S2) * t] (1) where I= the number of taxa appearing in the second year, E = the number of taxa lost between the censuses, S 1 = the number of taxa in year I, S2 = the number of taxa in year 2, and t =the number of years between the censuses (e.g., Magnuson eta!. 1994). Because

Magnuson et a!. ( 1994) demonstrated that fish species turnover rates were dependent on sampling interval, we calculated turnover for all possible combinations of sampling intervals from one to eleven years.

We used Jaccard's Coefficient of Similarity (J, Jaccard 1912) to determine whether turnover led to divergence in community composition over time. J was calculated as

J=a!(a+b+c) (2) 52 where a= the number of taxa common to both years, b= the number of taxa present only in year 1, and c = the number oftaxa present only in year 2. Possible values range from 0 to 1, with these extremes identifying communities with totally different vs. identical species lists, respectively. The calculations were performed on all possible combinations of samples for intervals ranging from one to eleven years.

We assessed whether communities diverged from initial conditions by determining whether there was a correlation between similarity and inter-census interval. This was assessed using a Mantel's test (Sokal 1979) to generate a null distribution of the relationship between time interval and community similarity. A similarity matrix was generated, representing the similarity between communities for each census interval. Similarity data were transformed (D = sqrt(1-J)) and standardized so that the mean equaled 0 and the

variance equaled one. A corresponding census interval matrix was also generated. The null

distribution was generated by repeatedly (10000 X) calculating the correlation between the

census interval matrix and a randomized version of the similarity matrix. The statistic was

then obtained by comparing the null distribution of correlations (based on the

randomizations) and the measured correlation between time interval and similarity. The

correlation was considered significant (not random) if the measured value was higher than

95% of the correlations obtained from the null distribution.

Error in turnover estimates

To determine the turnover rate attributed to sampling error, we conducted Monte

Carlo simulations for each lake based on the probability of detecting individual species. The 53 probability of detecting each species in each lake was calculated in two ways, assuming either a uniform or negative binomial distribution of species across time. For both methods, we assumed that the total species pool of each lake was present every year, but we occasionally failed to detect some because of chance. For the uniform distribution we assumed that each species had a different probability of detection, but that detection was independent of density. It was, however, dependent on the number of years the species was actually detected in the database. For the uniform distribution, we calculated the probability of detecting each species in each lake as

p =(number of years present)/(total years sampled) (3)

In these calculations, species density was not used to determine the probability of detection,

although density may have influenced it indirectly by influencing the number of years that the species was present.

For the negative binomial distribution, we assumed that the probability of detecting each species was based on its mean annual density. We were less likely to detect species

with low abundance. We calculated the probability of detecting each species in each lake as

p= r-cr+rnlkrk (4)

where m is the total number of individuals of the species in the six samples in the year, and k

is a fitted parameter. Because sampling protocol was similar through time, we employed the

number of individuals, rather than the density. In doing this we generalized the equation so it 54 could be used for sampling methods that census indeterminate areas of habitat, e.g., gillnets for fishes.

For each species, we used one of two methods to estimate 'k'. When a particular species was encountered in only one or two years, 'k' was obtained by solving

k * log(! + x/k) = log(N/f0 ), (5)

where N is the number of years, f0 is the number of years in which the species was not detected, and x is the mean number of individuals per year (Bliss and Fisher 1953). In other cases, k was obtained by solving the simpler form

where x is the mean number of individuals encountered each year, and s2 is the variance of number of individuals across years (Bliss and Fisher 1953).

For each lake, we classified species into several groups according to their probability of detection (equations 3 to 6): 0-1.0 (all species), 0-0.25 (rare species), 0.25-0.75

(intermediate), and 0.75-1.0 (common species), 0.25-1.0 (rare species removed). Species comprising each category were similar for both uniform and negative binomial distributions.

For each group, we used the calculated probability of detection of each species to construct a presence/absence matrix of all species within a lake during the 12-year period. Using the presence/absence matrix we calculated turnover using equation 1. The matrix for each group was reconstructed and turnover calculated 1,000,000 times to obtain a mean and distribution of turnover that we could attribute to sampling error (Fig. 1). 55 The influence of species density on persistence and transition probability

We used a regression analysis to determine whether appearances and disappearances

of species were related to species density. Rare species (less than 0.25 probability of

detection) were removed from the analyses. We tested whether species persistence (the

proportion of years a species was present) and probability of changing states from present to

absent and vice versa could be predicted from annual abundance of each species. Because

residuals were not distributed normally, we performed an arcsine transformation on the

dependent variables (persistence and transition probability). Annual abundance was log­

transformed to achieve linearity.

Richness Estimates

We determined long-term zooplankton species richness in two ways, because we

wished to compare the species actually observed with an estimate of the total species pool in

each lake. Cumulative richness was the total number of species actually detected in the

entire sampling record. The asymptotic richness was calculated to estimate the total species

pool, assuming an infinite number of samples were available. It is based on the Walford plot

(Ricker 1975) and was obtained by calculating where a plot of the mean cumulative richness

at sub-sample size t versus mean cumulative richness at sub-sample size t +I intersects the I: I

line. The Walford plot provides a simple method for estimating asymptotic richness that is relatively insensitive to sample size (Arnott et al. 1998). Calculations employed the total

number of available years of data ( 10-12 depending upon lake). 56 Results

A total of 36 crustacean zooplankton species were detected in the study lakes (Table

2). The mean annual richness of each lake ranged from 11.4 to 17.2 species, with an average of 15 (Table 3). Heney and Plastic Lakes had fewer species than the other lakes (Tukey a posteriori test P<0.05). Heney Lake is shallow and unstratified. Plastic Lake acidified during the record (Dillon et a!. 1987).

Throughout the 12 years, species richness in all lakes remained relatively constant.

The coefficient of variation (among years) for lakes ranged from 8 to 20% and averaged

13.2%, but the total number of species observed in each lake ranged from 20 to 31, with an

average of 25. Hence, an average of only 60% of the cumulative species pool was detected each year (Arnott eta!. 1998). With two exceptions, the asymptotic estimate predicted a

species pool greater than the cumulative species richness over the entire sampling period

(Table 3). If we assume that the communities were stable and not in transition to a new state,

then this indicates that long sampling records are necessary to detect all species that occur in

some lakes. Even after 12 years of sampling, we had recorded as little as 65% and an

average of 87% of the total species pool, based on an asymptotic estimate of species richness.

Only in Crosson and Red Chalk Lakes was 100% of the predicted species pool actually

sampled.

Species Turnover

Apparent species turnover rates between adjacent years averaged 16% (± 5% S.D,

range from 11 to 28 %) when all species were included in the analysis (Table 4, Fig. 2). 57 Apparent turnover rates did not differ among lakes, with the exception of Heney Lake, which had a higher apparent rate than all lakes except Dickie Lake (ANOVA main effect= lake,

Tukey a posteriori multiple comparison test P

(Levene's Test: P = 0.002). On average, lakes lost more species per year (2.3 species) than they gained (2.0 species) (paired T-test: P = 0.01).

Turnover usually altered species composition over time. Zooplankton communities in Blue Chalk, Chub, Dickie, Harp, and Plastic Lakes became more dissimilar as inter-census interval increased (Table 5). In contrast, sequential turnover in Red Chalk, Crosson, and

Heney Lakes did not produce different species assemblages. Rather, the same species appeared and disappeared in our annual censuses.

Turnover attributable to sampling error

The distribution of zooplankton in our annual composites was best approximated by the negative binomial distribution. A comparison of simulated and observed temporal distributions indicated that 58%± 12% of the taxa approximated the negative binomial distribution (Kolmogorov-Smirnov goodness of fit test for discrete data, P>0.05). Fewer of the taxa (41% ± 17%) approximated a uniform distribution (Kolmogorov-Smirnov goodness of fit test for discrete data, P>O.OS).

In most of our study lakes, apparent turnover was equivalent to the noise attributable to sampling or sample enumeration. Only in two of the eight lakes was the apparent turnover 58 greater than our estimate of sampling error. We are skeptical that the difference between apparent turnover and sampling turnover is a reasonable estimate of the actual turnover.

However, it is clear that apparent turnover is an over-estimate of the turnover that may be attributed to immigration and extinction of species.

Sampling error contributed most to apparent turnover of rare taxa. For species with a high probability of detection (>0.75) apparent turnover was greater than meau sampling turnover in 2 of 8 lakes for both distributional assumptions (uniform and negative binomial), but was only greater than 53 to 84% of the simulated values (sampling turnover) for these two lakes. For intermediate species (0.25-0.75 probability of detection) apparent turnover was higher than sampling error in five lakes assuming a negative binomial distribution and in three lakes assuming a uniform distribution. However, apparent turnover was greater than

90% of the simulated turnover values in two of the lakes assuming a negative binomial distribution and ranged from 50 to 82% assuming a uniform distribution. Likewise, turnover attributed to sampling error (hereafter called "sampling turnover") was higher than apparent turnover in five and three lakes for rare species assuming a negative binomial distribution or uniform distribution, respectively. When apparent turnover exceeded sampling turnover, the difference was marginal (apparent turnover> sampling turnover in 50 to 76% of the simulations). Therefore, apparent and sampling turnover were difficult to distinguish for rare

species. Turnover was most readily observable in the intermediate category; rare species

have high sampling turnover, and common species have low actual turnover. 59 The influence of species density on persistence and transition probability

When rare species (probability of detection <0.25) were removed from the analyses, rates of apparent species turnover were related weakly to the average abundance of species within each lake (calculated excluding non-occurrences). Only 44% of the variation in the proportion of years a species was detected could be explained by its abundance (P<0.001, R2

= 0.44). Species with high abundance tended to be present every year and those with low densities tended to be absent in several years. Similarly, the probability of transition from present to absent decreased with higher mean annual densities (P < 0.001), but density only explained 39% of the variation in the disappearance probability.

Species distributions among lakes and across years

The patterns of species presence/absence across years were bimodal; most species, being either rare (occurring in <25% of the years) or common (occurring in >75% of the

years, Fig. 3). In any year, a large group of species was detected in ailS-study lakes and

another large group was rare, occurring in fewer than two of the study lakes. The same

species did not comprise these groups in each year, however. Mean similarity (Jaccard's

Coefficient) among years of the species found in all years was 0.52 ± 0.14 S.D. Only two

taxa, Leptodiaptomus minutus and Diaphanosoma birgei, were found in every lake every

year. Of the 18 species that were permanent in at least one lake, only ten species were

permanent in more than 50% of the lakes. Thus, temporally common zooplankton were not

necessarily widespread across the study lakes. 60 Discussion

Zooplankton species composition commonly changes in lakes in response to experimental manipulation or environmental disturbance (e.g., Keller and Pitblado 1984;

Malley eta!. 1988). We have reported high apparent turnover rates of zooplankton in each of eight lakes that, with the exception of Plastic Lake (Dillon eta!. 1987), have not been obviously disturbed. This turnover was detected even though the study was not designed to detect turnover i.e. there was no particular focus on rare species or on disturbed lakes.

However, apparent turnover includes a large component of sampling error, i.e. the failure to detect species that were actually present. While this problem has a long history

(Fisher eta!. 1943) the growing interest in global biodiversity assessment has renewed the search for its resolution (e.g., Boulinier eta!. 1998). One approach has been to adapt jackknifing procedures developed for capture-recapture population studies to estimate species richness, extinction probabilities, and colonization rates (Nichols eta!. in press).

This technique is especially useful when variation in space exceeds that over time, and there are spatial replicates for each of two censuses. We have taken an alternate approach and have focussed on 'replicates' over time by using long-term data rather than spatial samples.

Our method provides a new and unique estimate of turnover attributable to sampling error.

By comparing turnover attributable to sampling with the apparent turnover rate, we discovered that much of the apparent turnover that we measured could be explained largely or entirely by sampling error. Even in the two lakes where the turnover attributable to sampling was less than the apparent turnover, the magnitude of this difference was small relative to the sampling turnover. High sampling turnover relative to apparent turnover 61 indicates that apparent turnover is likely to be a gross overestimate of actual turnover, especially for communities with numerous rare taxa.

Despite the difficulty in estimating turnover attributable to immigration and extinction of species, there is still value in quantifying apparent turnover. High apparent turnover rates, such as those measured for Dorset zooplankton, have important implications for assessments of biodiversity and interannual variability in community structure. Our

overall estimates of crustacean zooplankton species richness in the Dorset area were much

higher than the literature values, except when comparisons were based on the number of taxa per sample (Yan and Strus 1980; Keller and Yan 1991). Our species lists generated from a

single year's pelagic samples of crustacean zooplankton (six samples) underestimated the

total species pool based on 12 years of data by 33 - 50%. A single sample underestimated

the total species pool by 55%, on average. This information may enable us to estimate long­

term diversity in lakes based on limited samples. However, further studies are necessary to

determine whether these results are consistent in other regions and for other taxa. Additional

studies are required for littoral zooplankton (eg. Dumont and Segers 1996)

Apparent turnover rates were influenced by the inter-census interval. Turnover rates

calculated between distant years miss the frequent appearances and disappearances of the

same species (Fig. 4). Hence, our apparent turnover rates were much higher than those of

Browne (1981), who estimated that the zooplankton communities in central New York lakes

changed at a rate of 0.28%/yr over a 50 year period, an estimate many times lower than ours

(Fig. 2). This discrepancy may be attributable to the dependence of turnover rate on inter­

census interval, the denominator in equation 1, independent of changes in immigration or 62 extinction of taxa (Diamond and May 1977). Indeed, following Magnuson eta!. (1994), the estimate of annual turnover rate does decrease as the inter-census interval lengthens (Fig 3), primarily because the denominator (time) increases. If absolute, rather than the rate of, turnover is calculated (i.e. without the denominator), the dependence of turnover on inter­ census interval is almost eliminated. The dependence on time changes from strongly negative to slightly positive. That is, the similarity between communities separated by one year is only slightly more than the similarity between communities separated by several years. This may be the result of a core group of species being present every year and the remaining species being intermittently present and absent. If we extend our turnover estimates over a longer sampling interval by using the mean annual turnover rate that we estimated and divide it by the number of years between Browne's samples, we get a turnover of 0.30% (0.24% when rare species were removed); a rate close to that calculated by Browne.

We believe Browne under-estimated turnover rate by not having annual data on gains and losses of species.

Comparison of our apparent zooplankton turnover rates with those of other organisms is complicated because turnover rate is influenced by a number of factors. These include census interval, the number of census intervals included in the analysis, the occurrence of

rare species, and the magnitude of sampling error. In a review of turnover studies, Schoener

(1983) found that relative turnover tends to be lower for longer-lived, than shorter-lived

organisms. Considering one year census intervals and 12 intervals, our apparent turnover rate

estimates average 16%/yr (including all species), and 12% (rare species excluded). Ignoring

sampling error (as did Schoener), our crustacean zooplankton turnover falls somewhere 63 between Schoener's estimates for arthropods (25-60%) and terrestrial vertebrates (0.1-5% ).

If we consider the magnitude of sampling error, we realize that such a comparison among taxa is meaningless because much of the apparent zooplankton turnover can be attributed to

sampling error. High sampling error in zooplankton leads us to suspect that sampling error may also be important in other taxa. Therefore, Schoener's conclusion that turnover is

allometrically scaled may be an artifact caused by variation among species in sampling

turnover. Without quantifying the relative contribution of sampling error to apparent

turnover of other taxa, we cannot attribute differences in apparent turnover to differences in

life history characteristics. Standardized comparisons among taxa that determine the

magnitude of sampling error are required to verify the relationship between turnover rate and

size or generation time.

Explanations of species turnover

Appearances and/or disappearances of species from censuses can be caused by

several mechanisms, including some that are artifactual in nature. Artifactual mechanisms

include; 1) chance detection of rare species, and 2) variation in population densities above

and below detection limits because of changes in abundance or spatial distribution. The

actual mechanisms include; 1) aria! dispersal of zooplankton among lakes (Jenkins 1995), 2)

downstream dispersal of zooplankton between lakes, 3) human introductions of exotic and

native taxa (Yan and Pawson 1997), 4) emergence of zooplankton from sediment egg banks

(Hairston 1996), 5) extinctions resulting from environmental disturbances, and 6). chance

extinctions related to random variation in birth and death rates, or 7) unusual weather or 64 climatic conditions (Sternberger eta!. 1996).

Some of these mechanisms are probably not important drivers of turnover in our study lakes. The only exotic species that has invaded our study lakes (Bythotrephes cederstroemi) arrived in Harp Lake in 1993, i.e. after our study period (Yan and Pawson

1997). Hence, turnover was not influenced by the arrival of distant exotics.

Similarly, turnover is probably not related to direct human traffic among lakes. The

Spearman's rank correlation between apparent turnover and the number of dwellings (a surrogate of use and travel) on each lake is very low, at -0.03.

Sampling frequency is an important contributor to both apparent and sampling turnover in our data set. We compared turnover rates for the complete data set in 1978 and

1979 with our subset of 6 samples per year for six lakes (Table 6). On average, we found four additional species in the complete database in 1978 and three additional species in 1979.

Apparent turnover rates tended to be higher when using the complete database, except for

Dickie Lake (Table 6). Because of the increased number of samples in each year, sampling turnover (i.e. error) was lower for the complete database, averaging 5.8%, vs. 15.8% for the

6-sample subset. The combination of higher apparent turnover rates and lower sampling error resulted in a relatively large difference between apparent and sampling turnover in the five lakes between 1978 and 1979 when the complete database was considered. Therefore, it appears that our 6-sarnple subset of data provides conservative estimates of the importance of sampling error when compared with a more intensively sampled dataset available only in the first few years. This suggests that actual turnover of zooplankton may be quite common in 65 unmanipulated Shield lakes and that annual changes in species composition are more common than suggested by our calculations using the 6-sample subset of data.

Other mechanisms may be important drivers of species turnover in our study lakes.

Downstream dispersal of taxa from Blue Chalk to Red Chalk Lake could contribute to turnover. We did observe a larger difference between apparent turnover and potential

sampling turnover rates in Red Chalk (the only second order lake) than all other study lakes,

except Heney Lake.

Changes in acidity probably influenced annual turnover rates. Five of our study lakes

had a pH <6.0, the threshold commonly accepted for adverse effects of acidification on

zooplankton (e.g., Havens eta!. 1993). Apparent species turnover rates were higher in the 5

lakes with pH<6, than in the three non-acid lakes (P=O.OOl). This difference resulted primarily from differences in richness (species richness was lower in acid lakes than non-acid

lakes P

extinction rates were marginally greater in acid lakes (P=0.057).

Temporal changes in species composition provide further evidence of the potential

importance of environmental conditions as drivers of species turnover. Two lakes with

increasing dissimilarity in zooplankton composition (measured using Jaccard's coefficient of

similarity) underwent substantial chemical changes from 1979 to 1985. Plastic Lake

acidified from pH 5.8 to 5.6 (Dillon eta!. 1987) and TP concentrations fell by 40% in Dickie

Lake (P. Dillon, personal communication). Declines in pH in Crosson and Chub Lakes may

have occurred since pre-1850 (Hall and Smoll996), although there were no detectable

changes during this study period. 66 Implications of species turnover

The mechanism or mechanisms that regulate turnover has important implications for how lakes are monitored. If the abundance of many species fluctuates above and below limits of detection, more intense sampling programs would be required to understand community dynamics. Evidence of high dispersal from nearby lakes would necessitate a shift in conceptual focus from individual lakes to lake-regions. Rather than determining how individual lakes are affected by anthropogenic disturbances, it would be more important to determine how an entire region, the source of colonists, is affected and whether a diverse source of potential recruits is maintained. On the other hand, if the sediments are the primary source of colonists, factors that influence the probability of emergence, such as the rate of sedimentation, the mixing of sediments, and the duration of the disturbance to the lake would be important in determining prospects for recovery of damaged lakes.

The annual gains and losses of species that we observed resulted from a combination of several mechanisms. Our Monte Carlo simulations suggested that sampling error is a large component of the measured turnover, especially for rare species. However, not all apparent turnover was accounted for by our estimates of sampling error, suggesting that zooplankton species structure can change, even in relatively undisturbed lakes.

Acknowledgments

We thank Trevor Pawson, Robert Girard, Martyn Putter, and Paul Hanson for technical assistance and Bill Geiling for counting the zooplankton samples. Stanley Dodson, J arret

Fischer, Jen Klug, and Pat Soranno provided helpful reviews. Craig Stow, Conrad Lamond, 67 Keith Somers, and Steve Carpenter provided advice on analyses. The Dorothy Powers Grant and Eugene Lodewick Grant Scholarship Fund provided financial support for S.E.A.

Support for research was provided by the Ontario Ministry of the Environment and the North

Temperate Lakes Long-Term Ecological Research project funded by the National Science

Foundation, Grants BSR14330 and DEB9011660.

References

Arnott, S. E., Magnuson, J. J., Yan, N.D. Crustacean zooplankton species richness: single­ and multiple-year estimates. Can. J. Fish. Aqua!. Sci. 55:1573-1582.

Bliss, C. I. and Fisher, R. A. 1953. Fitting the negative binomial distribution to biological data. Biometrics 176-200.

Boulinier, T., Nichols, J.D., Sauer, J. R., Hines, J. E., and Pollock, K. H. 1998. Estimating species richness: the importance of heterogeneity in species delectability. Ecology 79:1018-1028.

Browne, R. A. 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. J. Biogeogr. 8: 75-83.

Chen, C. Y. and Folt, C. L. 1996. Consequences of fall warming for zooplankton overwintering success. Lirnnol. Oceano gr. 41:1077:1086.

Colbourne, J. K. and. Hebert, P. D. N. 1996. The systematics of North American Daphnia (Crustacea: Anomopoda): a molecular phylogenetic approach. Phil. Trans. Roy. Soc. Lond. B 351:349-360.

DeMelo, R., and Hebert, P. D. N. 1994. Allozymic variation and species diversity in North American Bosminidae. Can. J. Fish. Aquat. Sci. 51:873-880.

De Stasio Jr., B. T., Hill, D. K., Kleinhans, J. M., Nibbelink, N. P., and Magnuson, J. J. 1996. Potential effects of global climate change on small north-temperate lakes: Physics, fish, and plankton. Lirnnol. Oceanogr. 41:1136-1149.

Diamond, J. M. and May, R. M. 1977. Species turnover rates on islands: dependence on census interval. Science 197: 266-270. 68 Dillon, P. J., and Molot, L.A. 1996. Long-term phosphorus budgets and an examination of a steady-state mass balance model for central Ontario lakes. Wat. Res. 30:2273-2280.

Dillon, P. J., Reid, R. A. and de Grosbois, E. 1987. The rate of acidification of aquatic ecosystems in Ontario, Canada. Nature 329: 45-48.

Dobson, A. P., Rodriguez, J.P., Roberts, W. M., and Wilcove, D. S. 1997. Geographic distribution of endangered species in the United States. Science 275:550-553.

Dodson, S. I. 1981. Morphological variation of Daphnia pulex Leydig (Crustacea: Cladocera) and related species from North America. Hydrobiol. 83: 101-114.

Dumont, H.J. and Segers, H. 1996. Estimating lacustrine zooplankton species richness and complementarity. Hydrobiol. 341: 125-132.

Fisher, R. A., Corbet, A. S., and Williams, C. B. 1943. The relation between the number of species and the number of individuals in a random sample of an animal population. J. Animal Ecol. 12:42-58.

Hairston Jr., N. G. 1996. Zooplankton egg banks as biotic reservoirs in changing environments. Limnol. Oceano gr. 41:1087-1092.

Hall, R.I. and Smol, J.P. 1996. Paleolimnological assessment of long-term water-quality changes in South-Central Ontario lakes affected by cottage development and acidification. Can. J. Fish. Aquat. Sci. 53:1-17.

Harper, J. L. and Hawksworth, D. L. 1994. Biodiversity: measurement and estimation. Phil. Trans. R. Soc. Lond. B. 345: 5-12.

Havens, K. E., Yan, N.D., and Keller, W. 1993. Lake acidification: effects on crustacean zooplankton populations. Environ. Sci. Techno!. 27: 1621-1624.

Hebert, P. D. N. and. Finston, T. L. 1997. Taxon diversity in the genus Holopedium (Crustacea: Cladocera) from the lakes of eastern North America. Can. J. Fish. Aquat. Sci. 54: 1928-1936.

Hebert, P. D. N. and Finston, T. L. 1997. A taxonomic re-evaluation of North American Daphnia 3. The Daphnia catawba complex Can. J. Zoo!. 75:1254-1261.

Hecnar, S. J. and M'Closkey, R. T. 1996. Regional dynamics and the status of amphibians. Ecology 77:2091-2097.

Heywood, V.H. 1995. Global Biodiversity Assessment. Cambridge University Press, Cambridge. 69 Jaccard, P. 1912. The distribution of the flora in the alpine zone. New Phytol. 11:37-50.

Jenkins, D. G. 1995. Dispersal-limited zooplankton distribution and community composition in new ponds. Hydrobiol. 313/314:15-20.

Keller, W. and Pitblado, J. R. 1984. Crustacean plankton in northeastern Ontario lakes subjected to acidic deposition. Water Air and Soil Pollut. 23:271-291.

Keller, W. and Yan, N.D. 1991. Recovery of crustacean zooplankton species richness in Sudbury area lakes following water quality improvements. Can. J. Fish. Aquat. Sci. 48: 1635-1644.

MacArthur, R. H., and Wilson, E. 0. 1967. The Theory of Island Biogeography. Monographs in Population Biology.

Magnuson, J. J., Benson, B. J. and McLain, A. S. 1994. Insights on species richness and turnover from long-term ecological research: fishes in north temperate lakes. Am. Zoo!. 34: 437-451.

Malley, D.F., Chang, P.S.S., Findlay, D.L., and Linsey, G.A. 1988. Extreme perturbation of the zooplankton community of a small Precambrian Shield lake by the addition of nutrients. Verh. Internat. Verein. Limnol. 23:2237-2247.

McCormick, P. V. and Cairns Jr, J. 1990. Microbial colonization dynamics in temporary aquatic systems. Hydrobiol. 196: 229-253.

Nichols, J.D., Boulinier, T., Hines, J. E., Pollock, K. H., and Sauer, J. R. Estimating rates of local species extinction, colonization and turnover in animal communities. Ecol. Appl. In Press.

Ricker, W. E. 1975. Computation and interpretation of biological statistics of fish populations. Department of the Environment Fisheries and Marine Service. Bulletin 191.

Schoener, T. W. 1983. Rate of species turnover decreases from lower to higher organisms: a review of the data. Oikos 41: 372-377.

Schoener, T. W. and Spiller, D. A. 1987. High population persistence in a system with high turnover. Nature 330: 474-477.

Sokal, R. R. 1979. Testing statistical significance of geographic variation patterns. Syst. Zoo!. 28:227-232. 70 Sternberger, R. S., Herlihy, A. T, Kugler, D. L., and Paulsen, S. G. 1996. Climatic forcing on zooplankton richness in lakes of the northeastern United States. Limnol. Oceanogr. 41:1093-1101.

Taylor, D. J. and. Hebert, P. D. N. 1994. Genetic assessment of species boundaries in the North American Daphnia longispina complex (Crustaceana: Daphnidae). Zoo!. J. Linnean Soc. 110:27-40.

Williamson, C. E., Sternberger, R. S., Morris, D.P., Frost, T. M., and Paulsen, S. G. 1996. Ultraviolet radiation in North American Lakes: Attenuation estimates from DOC measurements and implications for plankton communities. Limnol. Oceanogr. 41:1024-1034.

Yan, N.D. and Strus, R. 1980. Crustacean zooplankton communities of acidic, metal­ contaminated lakes near Sudbury, Ontario. Can. J. Fish. Aquat. Sci. 37: 2282-2293.

Yan, N.D. and Pawson, T. W. 1997. Changes in the crustacean zooplankton community of Harp Lake, Canada, following the invasion by Bythotrephes cederstroemi. Freshwat. Bioi. 37:409-425.

Yan, N.D., Keller, W., Somers, K. M., Pawson, T. W. and Girard, R. E. 1996. The recovery of crustacean zooplankton from acid and metal contamination: comparing manipulated and reference lakes. Can. J. Fish. Aquat. Sci. 53:1301-1327. Table 1: Morphometric and chemical data for the study lakes during the period 1978 to 1989. Values represent the mean of all samples taken during the ice-free season (generally biweekly to monthly) over the twelve-year study period. Standard deviations for chemical data are in parentheses. Lake Latitude Longitude Area Mean Maximum Alkalinity Conductivity DOC pH TP

(N) (W) (ha) Depth (m) Depth (m) mg/L CaC03 (J1S) (mg!L) (Jlg/L)

0 Blue Chalk 45° 11' 78 56' 52.4 8.5 23.0 4.09(0.34) 28.92 (0.67) 1.79 (0.20) 6.65 (0.15) 6.30 (0.87)

0 0 Chub 45 13' 78 59' 34.4 8.9 27.0 0.81 (0.23) 27.87 (0.65) 4.67 (0.28) 5.62 (0.14) 10.39 (1.42)

0 0 Crosson 45 05' 78 02' 56.7 9.2 25.0 0.55 (0.18) 25.99 (0.46) 4.03 (0.23) 5.58 (0.11) 11.00 (1.42)

0 0 Dickie 45 09' 79 05' 93.6 5.0 12.0 1.15 (0.41) 28.29 (0.77) 4.97 (0.33) 5.86(0.17) 11.38( 2.07)

0 0 Harp 45 23' 78 08' 71.4 13.3 37.5 3.29 (0.20) 34.40 (0.65) 3.81 (0.24) 6.29 (0.10) 7.52 (1.08)

0 0 Heney 45 08' 79 06' 21.4 3.3 5.8 0.51 (0.22) 25.03 (0.81) 2.82 (0.26) 5.81 (0.17) 7.04 (1.38)

0 0 Plastic 45 11' 78 50' 32.1 7.9 16.3 0.52 (0.19) 22.48 (0.57) 2.22 (0.14) 5.74 (0.13) 5.96 (0.98)

0 Red Chalk 45° 11' 78 56' 32.1 16.7 38.0 3.31 (0.24) 29.39 (0.40) 2.45 (0.19) 6.33 (0.16) 5.06 (0.71)

_, 72 3 Table 2: Density (number • m- ) (first row) and persistence (the percent years that each species was detected) (second row) of zooplankton in Dorset lakes. Values for rare species (less than 25% probability of detection assuming a negative binomial distribution) are in boldface. species name Blue Chub Crosson Dickie Heney Harp Plastic Red Chalk Chalk Acroperus harpae 1.3 1.0 8 11 Alana guttata 0.1 0.2 0.9 1.1 8.9 8 8 8 17 22 Bosmina longirostris 320 458 656 736 1140 2382 1008 158 100 92 100 100 100 100 100 100 Ceriodaphnia lacustris 0.6 0.8 7.4 9.4 0.7 8 8 22 20 8 Chydorus bicornutus 0.2 8 Chydorus sphaericus 5.1 0.9 6.9 2.9 5.6 787 4.5 0.8 50 17 58 33 44 100 10 17 Daphnia ambigua 0.7 51.2 12.5 345 1.4 521 1.3 8 92 33 83 17 90 25 Daphnia catawba 45.2 130 1174 1.4 0.7 6.5 17.2 33 50 67 11 8 10 25 Daphnia dubia 362 0.5 4.5 10.1 0.9 100 100 8 17 25 8 92 Daphnia galeata mendotae 1211 12.7 598 83.1 11.3 222 13.5 675 100 67 100 67 44 100 20 100 Daphnia longiremis 384 804 6.0 10.7 1.2 128 84 100 100 17 11 8 40 100 73 Table 2 (continued): Density and persistence species name Blue Chub Crosson Dickie Heney J-larp Plastic Red Chalk Chalk Daphnia pulex 296 0.3 0.6 3.2 267 110 92 8 8 25 40 92 Daphnia retrocurva 12.9 313 12.4 1340 1315 36.7 4.8 25 100 33 100 100 92 42 Eubosmina coregoni 0.5 2.6 1.5 1.4 17 25 17 11 Eubosmina tubicen 9.0 558 313 1601 4701 504 3861 76.9 42 75 75 100 100 100 100 100 Holopedium gibberum 120 659 1318 1303 190 320 1564 138 100 100 100 100 89 100 100 100 Leptodora kindtii 5.6 1.7 6.3 7.9 1.5 2.3 33 33 33 50 25 42 Ophryoxus gracilis 0.3 8 Polyphemus pediculus 1.0 1.7 1.0 3.7 22.6 1.2 25 8 11 33 80 25 Sida crystallina 1.3 0.4 7.6 0.6 62.0 17 8 33 8 100 Chydorus globosus 2.1 11 Diaphanosoma birgei 188 241 689 511 2619 723 303 263 100 100 100 100 100 100 100 100 Leptodiaptomus ashlandi 2.0 11 Leptodiaptomus minutus 1064 1339 1783 2959 5075 416 2510 442 100 100 100 100 100 100 100 100 74 Table 2 (continued): Density and persistence species name Blue Chub Crosson Dickie Heney Harp Plastic Red Chalk Chalk Skistodiaptomus oregonensis 249 0.8 298 0.9 3.4 0.5 77.3 100 8 100 17 33 10 100 Leptodiaptomus sicilis 1.3 17.0 8 92 Epischura lacustris 62.7 71.0 45.0 26.6 22.5 39.4 3.2 4.8 100 92 83 75 44 92 30 50 Senecella calanoides 0.9 3.1 80.5 8 11 100 Cyclops bicuspidatus 315 2.9 620 21.5 2.9 196 10.2 243 thomasi 100 42 100 75 22 100 50 100 Cyclops scutifer 322 228 118 220 1.5 0.3 299 228 100 100 100 100 11 17 100 100 Cyclops vernalis 4.4 3.7 2.8 16.5 14.3 17 25 25 50 58 Eucyclops agilis 1.4 1.5 17 11 Eucylops speratus 0.7 0.4 8 8 Mesocyclops edax 179 146 303 862 540 32.6 462 47.4 100 100 100 100 100 92 100 100 Orthocyclops modestus 6.4 11.5 0.8 1.1 12.9 42 42 11 17 40 Tropocyclops prasinus 62.4 817 358 1081 1350 577 153 41.6 mexicanus 92 100 100 100 100 100 90 100 75 Table 3: Species richness of Dorset study lakes. Mean annual richness was calculated over

12 years. The coefficient of variation over time is given in parentheses. Cumulative richness

is the total number of species found during entire sampling period. The asymptotic richness

was calculated based on a Walford plot of cumulative richness at year t versus cumulative

richness at year t + 1.

Long-term Richness

Lake Mean Annual Cumulative Asymptotic

Blue Chalk 15.2 (13.7) 23 24

Chub 15.0 (13.9) 26 40

Crosson 15.2 (8.4) 24 24

Dickie 16.2 (15.8) 31 38

Harp 16.0 (10.0) 25 30

Heney 11.4 (20.5) 26 33

Plastic 12.3 (14.9) 20 21

Red Chalk 17.2 (8.6) 24 24

mean 14.8 24.9 29.2

S.D. 1.9 2.9 7.1 Table 4: Mean annual turnover rates(%) of zooplankton species from 1978 to 1989 in 8 Dorset area lakes. Species were divided into five categories based on either the negative binomial distribution or a uniform distribution: all species, all except rare species, rare species with a probability of detection <0.25, intermediate species with a probability of detection 0.25 to 0.75, and common species with a probability of detection >0.75. For categories other than 'all species', species composition varies slightly depending on the distribution used. Turnover attributable to sampling was calculated using Monte Carlo simulations and assuming either a negative binomial distribution of species across time or a uniform distribution. Boldface turnover rates indicate where apparent turnover was greater than sampling turnover. Negative Binomial Distribution

All Species Rare removed Rare Intermediate I Common ' Lake Apparent Sampling Apparent Sampling Apparent Sampling Apparent Sampling J Apparent Sampling Blue 11.4 13.1 8.2 10.3 54.5 65.5 68.3 60.0 I 1.1 2.5 Chalk Chub 15.8 16.9 11.8 12.1 69.1 81.0 68.0 53.2 I 2.6 4.02 Crosson 12.6 13.6 9.7 10.8 72.7 67.6 57.2 67.7 2.5 2.6 Dickie 19.6 22.4 14.9 16.8 86.9 84.5 I 46.5 48.6 2.6 2.9 I Harp 11.4 14.9 7.5 11.6 75.8 75.7 28.6 40.7 1.6 2.2 Heney 27.8 22.9 14.5 12.6 87.5 86.6 65.4 62.6 3.2 2.0 Plastic 15.7 17.3 14.0 15.7 55.6 44.6 48.7 46.8 4.2 3.9 Red 12.7 11.8 11.1 10.5 45.5 46.1 78.1 49.6 1.9 2.2 Chalk

....a- Table 4 (continued): Mean annual turnover rates(%) of zooplankton

Uniform distribution

All Species Rare removed Rare Intermediate Common Lake Apparent Sampling Apparent Sampling Apparent Sampling Apparent Sampling Apparent Sampling Blue 11.4 11.6 6.9 7.6 54.5 65.6 65.8 65.8 i 1.1 1.2

.1 Chalk Chub 15.8 17.1 10.6 11.6 45.5 70.7 65.6 55.8 2.2 2.2 ! Crosson 12.6 13.8 7.4 8.8 72.7 65.7 54.2 63.8 ' 0.8 1.2 Dickie 19.6 21.2 13.0 13.5 86.9 85.0 42.2 49.6 I 1.0 1.5 Harp 11.4 11.8 6.2 6.7 75.8 76.0 56.1 61.9 I 2.4 2.3 I Heney 27.8 23.1 11.3 11.1 87.5 64.6 59.2 64.6 I 1.7 1.3 I Plastic 15.7 17.1 10.8 13.2 70.4 70.4 55.9 64.9 ! 3.1 3.7 Red 12.7 10.1 5.9 5.5 45.5 47.4 74.2 62.8 I 0.9 1.1 Chalk

.... 78 Table 5: Summary statistics from Mantel's test for Jaccard's Similarity versus the number of years between sampling dates. Years between sampling dates ranged from 1 to 11. In lakes

(boldface) with P-values less than 0.05 the species composition became more dissimilar as the interval between sampling increased.

Lake n Pearson's ·Significance level

Coefficient

Blue Chalk 11 0.25 0.04

Chub 11 0.40 <0.01

Crosson 11 0.04 0.35

Dickie 11 0.35 O.ol

Harp 11 0.26 0.03

Heney 8 -0.07 0.64

Plastic 9 0.41 O.ol

Red Chalk 11 0.16 0.13 79 Table 6: Comparison of species richness and turnover rates for complete database and six­ sample subset. Apparent turnover rate represents the observed change in species composition between 1978 and 1979.

Complete Database Six-Sample Subset

Lake no. 1978 1979 Apparent 1978 1979 Apparent

samples richness richness turnover richness richness turnover

rate rate

Blue Chalk 18 19 18 13.5 18 15 9.1

Chub 21 22 17 17.9 19 15 17.6

Dickie 27 30 23 17.0 21 21 19.0

Harp 20 24 23 10.6 19 19 10.5

Red Chalk 18 21 22 11.6 19 19 10.5

Mean 20.8 23.2 20.6 14.1 19.2 17.8 13.3 80 Figure Captions

Figure 1. Flow chart indicating steps of Monte Carlo simulation to estimate potential

sampling error associated with apparent species turnover. The right hand column

provides hypothetical examples associated with each step.

Figure 2. Apparent turnover rates between adjacent years for Blue Chalk (Be), Chub (Ch),

Crosson (Cr), Dickie (Dk), Harp (Hp ), Heney (Hn), Plastic (PI), and Red Chalk (Rd).

Box plots indicate the median turnover rate for each lake. The hinge, or outer edge of

each box indicates the upper and lower quartile and the vertical lines indicate the range in

the data. The notches, surrounding the medians provide a rough measure of the

significance of differences between values. If the notches do not overlap, the medians are

roughly significantly different at a 95% confidence level. In some cases, the notches

protrude beyond the hinges. Asterisks represent values that are outside the inner fence

(lower hinge - l.5*hingespread or upper hinge+ 1.5* hingespread). Open circles

represent points that are outside the outer fences (lower hinge - 3*hingespread or upper

hinge + 3* hingespread).

Figure 3: Mean temporal distributions of crustacean zooplankton in the 8 study lakes.

Because each lake had a similar distribution pattern, we plotted the average distribution

across the 8 study lakes. The darkly shaded portion of the '76-100' category indicates the 81 percent species that were detected every year (100% of the years). Error bars are one

standard error of the mean, based on 8 lakes.

Figure 4: Mean apparent turnover rates based on all possible combinations of each sampling

interval. Rare species ( < 25% probability of detection) were not included in this analysis.

The dashed line represents the mean apparent turnover rate between adjacent years for

all 8 study lakes. The solid circles indicate the turnover rate as the number of years

between samples increases (T=lOO*(I+E)/(S 1+Sz)*t). The open circles represent the

turnover rate if the number of years between samples is not accounted for

(T=l00*(I+E)/(S1+S 2)). The solid line is the mean annual turnover rate divided by an

increasing number of years (mean turnover/t). Error bars represent one standard error of

the mean. 82

Select a lake

Calculate probability Species I Species 2 Species 3 etc. of detection for each 0.95 0.68 0.99 species

Year

Determine I 1 1 I 1 1 1 presence/absence for 1 study duration based 1 1 1 1 on probability of 1 1 1 1 detection 1 1 1 1 1 1 1 1 1 1 1

Calculate turnover between adjacent years

T =100* (#gained+# lost)/(S 1 + S2)

Average turnover rates across study period Repeat one million times

Distribution and mean of sampling turnover Jl nOO~Ono I Sampling turnover

Fig. 1 83

40

,-_ 30 ~ '-" ...... il..) ro ~ 20 1-< il..) 0 > 0 :::::: 1-< • ;:::::1 10 ~ ~ • 0 Be Ch Cr Dk Hp Hn Pl Rd Lake

Fig. 2 84

0 0 T""" coI 1'--

~ L!) -0 1'-- I I U) T""" -lo... ~ L!) ctl 0 (!) 0 > """""' -0 0 s:: (C) L!) -(!) I oil (,) co lo... ~ C\1 (!) c.

L!) C\1 I T"""

0 0 0 co C\1

(%) sa1oads JO lUaoJad ..--< ..--< I I 0 I ..--< I 0\ "'~ 00 -c. I s I ~ 00. t-- ::::: I ~ ~ l \0 ~ ~ I ~ I ~ "'on I trl iZ I "';.., ~ ~ ..;- >- r

86 CHAPTER THREE

LONG-TERM SPECIES TURNOVER AND RICHNESS ESTIMATES: A

COMPARISON AMONG AQUATIC ORGANISMS3

Abstract

We compared annual species turnover rates for three groups of aquatic organisms in undisturbed north temperate lakes. Apparent turnover rates (i.e. measured turnover rates) were high, averaging 18% for phytoplankton, 16% for zooplankton, and 20% for fishes. The potential contribution of sampling error to apparent turnover was high because some species are not detected in sampled populations. Sampling error was highest for fishes, followed by

zooplankton, then phytoplankton. Because of high sampling error, we could not be certain

that any of the apparent turnover could be attributed to actual immigration and extinction of

species. Based on life history characteristics and dispersal abilities, we expected

phytoplankton to have higher turnover rates than zooplankton, which would have higher

turnover rates than fishes. Results were contrary to our expectations; apparent turnover was

high and similar for each of the taxonomic groups. Our comparison of apparent turnover

rates, however, was problematic because sampling error could account for much of the

apparent turnover. Because the turnover that could be attributed to sampling error was so

high, it must be taken into consideration when assessing species turnover. We have

developed a new and unique method for quantifying potential sampling error.

3 Arnott, S. E., J. J. Magnuson, N.D. Yan, D. Findlay 87 Introduction

Efforts to understand the extent to which human activities influence species composition must be accompanied by an assessment of background changes in relatively undisturbed systems. Changes in species composition have frequently been measured as species turnover; the percent change in species composition between years (Diamond 1969).

Species turnover rates have been estimated for a number of organisms and habitats; birds

(Diamond 1969, Terborgh and Faaborg 1973, Hunt and Hunt 1974, Wright 1985), mammals

(Brown 1971), terrestrial invertebrates (Simberloff and Wilson 1969, Schoener and Spiller

1987), plants (Rusch and van der Maarel 1992, Milberg and Hansson 1993, Margules eta!.

1994, Rogers and Morrison 1994), and aquatic organisms (McCormick and Cairns 1990,

Magnuson et al. 1994, Arnott eta!. 1998b). In general, observed species turnover rates were I,' I: I. high, suggesting that changes in community composition are common. Comparisons of I ! turnover rates among taxonomic groups have indicated that the extent of change in I, community composition may be allometrically scaled (Schoener 1983). However, the

generality of this result is inconclusive because many studies have not accounted for error

associated with sampling populations.

Species turnover is comprised of two components: 1) Actual immigration and

extinction of species, and 2) sampling errors associated with not detecting species that are

present. Sampling error may influence the perception of species turnover for groups of

organisms because it results from underestimating the species richness at one or more

community censuses. Failure to detect species that are present results in an underestimate of

species turnover because species that have immigrated but exist at low densities are not 88 detected. Species turnover may be overestimated because failure to detect a species at low density can be mistaken for an extinction. Sampling error depends on the heterogeneity of organisms (both spatial and temporal), the catchability of organisms (i.e., nets catch different species with different efficiences), and the sampling intensity (Magnuson et al. 1994, Amott et al. 1998b). The result is that each species has a different probability of detection (Arnott et al. 1998b, Boulinier et al. 1998) and therefore contributes differently to sampling error.

Furthermore, previous studies have indicated that sampling error is high for some groups of taxa and may compromise our ability to quantify changes in species composition that result from the immigration and extinction of species (Magnuson et al. 1994, Arnott et al. 1998,

Nichols et al. 1998). Therefore, assessments of species turnover rates must take sampling error into account.

Apparent species turnover rates are influenced not only by sampling error, but also by environmental conditions and individual life history characteristics. Turnover rates are influenced by individual species responses to changes in environmental conditions (e.g. weather induced changes) and stochastic proc<;sses (vagaries in birth and death rates). A result of this is that organisms in similar habitats but different locations may have different apparent turnover rates (e.g., Toft and Schoener 1983). Apparent turnover rates also vary among groups of organisms (Schoener 1983). These differences have been correlated with life history characteristics such as generation time, population growth rates, and body size

(Schoener 1983, Magnuson et al. 1994). The interpretation of these results, however, is problematic when sampling error is not considered. 89 Lakes have several advantages for evaluating the turnover rates of organisms. A habitat can be easily defined because its boundaries are generally at the land-water interface for many aquatic organisms. Lakes are isolated habitats, much like islands. Based on island biogeography theory, turnover can be expected to vary as a function of island characteristics, such as degree of isolation and species characteristics, such as longevity and generation time.

Isolation within a particular habitat varies across taxonomic group, depending on the organism's vagility. For example, isolation of lakes is highest for fishes, which generally disperse very poorly among separate water bodies (Barbour and Brown 1974, Tonn et al.

1990, Magnuson eta!. 1998). In contrast, phytoplankton are thought to be are highly vagile.

There may be a continuous 'seeding' of lakes by phytoplankton from nearby lakes (Round

1971 ). Therefore, we would expect turnover rates to vary among groups of organisms, based

on their dispersal probabilities.

Here we compare turnover rates among three different groups of aquatic organisms;

phytoplankton, crustacean zooplankton, and fishes. Because differences in turnover rates

among groups of taxa may be confounded by sampling error, we estimated apparent turnover

based on observations of presence/absence and turnover that could potentially be attributed to

sampling error. Based on dispersal abilities and life-history characteristics, we expected

phytoplankton to be more responsive to environmental changes and therefore, turnover rates

should be high. In contrast, we expected turnover rates for fishes to be low due to low

dispersal rates and long generation times and lifespans. We expected zooplankton to have

intermediate turnover rates. 90 To determine if our estimates of turnover provided results consistent with other methods, we compared species turnover with two indices of similarity; Jaccard's coefficient, based on presence/absence and Morisita's index, based on relative abundance. Annual estimates of species richness were compared with long-term cumulative and asymptotic richness estimates (Amott et al. 1998a) to determine possible influences of compositional change on measures of biodiversity.

Because long-term, methodologically consistent databases are imperative for these analyses, but also rare, we compared three groups of organisms from three different lake regions. Therefore, taxonomic group and lake region were perfectly confounded. We believe, however, that life history differences and dispersal abilities among taxonomic groups surpass differences among lake regions. In particular, characteristics expected to be associated with dispersal are similar among regions (Table 1). Human activity, arguably the most important factor in the transport of aquatic organisms, differed among regions, but in an opposite manner to dispersal abilities associated with taxonomic groups. That is, fishes, which were expected to have low dispersal rates, were from a region where human-assisted

dispersal is probably the highest. Phytoplankton, which were expected to have high dispersal rates, were from a region with little human activity. Therefore, differences that we observed

among taxonomic groups were probably conservative estimates because of regional

differences.

Our emphasis in this paper is on a comparison of these three different groups of

aquatic organisms. Similar, but more detailed analyses for zooplankton have been conducted

previously (Amott et al. l998a, Amott et al. 1998b). Likewise, species turnover rates have 91 been calculated for LTER fishes (Magnuson eta!. 1994). We extend these analyses to include longer study duration and additional analyses.

Methods

Study sites

The study lakes are from three long-term study sites; the North Temperate Lakes

Long-Term Ecological Site in Wisconsin (NTL-LTER) (Magnuson eta!. 1984, Magnuson et

al. 1990), the Dorset Environmental Science Centre (Dorset) in south central Ontario (Dillon

et al. 1993), and the Experimental Lakes Area (ELA) in northwestern Ontario (Reeky et a!.

1994). All three sites are located in the upper Great Lakes region in forested catchments

(Fig. 1). The three sites differ in hydrology (surface water flow is dominant in the two ·:.:: Ontario sites whereas ground water flow is dominant in the Wisconsin lakes), forest ''"... ,

composition (spruce and jack pine at the ELA and mixed conifers and hardwoods at Dorset

and NTL-LTER), extent of residential and road development (high at Dorset and NTL-

LTER, very limited at ELA), and recent disturbances (fires and local clear cutting are

common at the ELA, but rare at the other sites) (Webster eta!. ms, Table 2).

Data used in this study minimize sampling biases. They are some of the most

extensive and methodologically consistent databases available in North America.

Throughout the study (duration ranges from 10 to 16 years) sampling and counting protocols

have remained internally consistent at each site. Lakes were sampled at the same location,

using the same gear, deployed in the same manner. Most importantly, the same person (D.

Findlay for ELA phytoplankton, Dr. W. Geiling for Dorset zooplankton) counted all 92 phytoplankton and zooplankton samples for each lake. In addition, multiple phytoplankton and zooplankton samples were taken each year, thereby minimizing variation resulting from seasonal changes in species composition.

Phytoplankton sampling- ELA Phytoplankton samples were collected throughout the ice-free season at a station located over the maximum depth of the lake. An integrating sampler was used to collect water from the epilimnion, metalimnion, and hypolimnion (Shearer 1978). Depth boundaries for each stratum were predetermined from light and temperature profiles. A 125-mL aliquot of lake water was preserved in Lugol's solution. Ten-mL aliquots of the preserved sample it. IP ' ::: were gravity settled for 24 hours then counted on an inverted microscope. All counts used :I the Utermohl technique as modified by Nauwerck (1963). Samples from each depth were counted separately then results were pooled so that each depth strata was hypsometrically weighted. For each lake we used data from 10- 15 years. Phytoplankton generally were identified to species. However, in some cases taxa were combinded to the genus level when identification to the species level was less certain.

Zooplankton sampling - Dorset

Zooplankton were sampled from 1978 to 1989 during the ice-free season.

Zooplankton were collected using a metered DRC (Dorset Research Centre) plankton net

(McQueen and Yan 1993) at a single station located over the maximum depth of the lake.

The net was hauled from several depths such that the sample composition was representative of the lake volume in each stratum (Yan et al. 1996). Samples were preserved with 6% 93 buffered formation. A minimum of 250 crustacean zooplankton was counted in each sample

(Allen et al. 1994). All Cladocera and Copepods were identified to species. However, we

combined counts of some species to allow for changes in nomenclature over the course of the

study (see Arnott et al. 1998a for details).

Fish sampling- North Temperate Lakes LTER site

Lakes were sampled annually using a combination of six methods: 1) 18 night beach

seine hauls (Benson and Magnuson 1992); 2) six, 24-hr inshore fyke net sets (Lyons and

Magnuson 1987); 3) 30 crayfish trap sets (unpublished LTER protocol), 4) one night of

electroshocking from a boat in four littoral areas (Lyons and Magnuson 1987); 5) two 24- I. ::.1i.f!l hour trammel net sets on the bottom across the thermocline (unpublished LTER protocol); n :

and 6) two 24-hr sets with 7 vertical gill nets ranging from 19 to 89 mm stretch mesh

(Rudstam and Magnuson, 1985). This variety of fish sampling equipment was chosen to

account for variation among species in habitat use and gear susceptibility. The same sites

were sampled each year. Annual sampling w;:ts conducted in late July and early August,

typically two lakes per week in random order. For additional details on the study site and

sampling protocol see Magnuson et al. (1994).

Data analysis

All analyses were based on summarized presence/absence or average abundance data

at the annual scale. Because different numbers of phytoplankton and zooplankton samples 94 were taken each year, we standardized the annual composite to six monthly samples, which

were taken from May to October. For fishes we determined presence/absence and total

number of individuals caught from total catch by all 6 gear-types.

Richness calculations

Long-term richness was estimated based on the entire sampling record, using three

methods. Mean annual richness was calculated as the average of the total number of species

detected in all samples for each year. Cumulative richness was the total number of species

detected during the entire sampling record. The asymptotic richness was calculated to

estimate the total species pool, given an infinite number of samples. It is based on the

Walford plot (Ricker 1975) and was obtained by calculating the point where the slope of the

mean cumulative richness at sub-sample size t versus mean cumulative richness at sub­

sample size t+ I intersects the 1: !line. The Walford plot provides a simple, easy-to-use

method for estimating asymptotic richness and is less sensitive to sample size than other

methods (Arnott et al. 1998a). 95

Species Turnover calculations

Apparent turnover

Apparent species turnover (between adjacent years) was estimated for each lake as

T=100* (I+E) (S,+S,l*t (1) where I= the number of new species, E = the number of species lost, S 1 = the number of species in year 1, S2= the number of species in year 2, and t = the number of years between samples (Diamond 1969).

Sampling turnover

Some species contribute to the apparent turnover because they are not detected every year, even though they are present. To determine the turnover rate attributable to this sampling error, we conducted Monte Carlo simulations developed for zooplankton (Arnott et al. 1998b). We calculated the turnover rate that would result if all species were continuously present, but were not detected every year. The probability of detecting each species was calculated assuming a negative binomial distribution across years (Arnott et al. 1998b ). This

assumption is justifiable because of the tested models (uniform, poisson, and negative

binomial) most species approximated a negative binomial distribution (58 % ± 12% of the

zooplankton; 78 % ± 11% of the fishes; and 86% ± 14% of the phytoplankton

(Kolmogorov-Smirnov goodness of fit test for discrete data, P>0.05). 96 The probability of detecting each species was calculated as

p = 1-(l+m/krk (2) where p is the probability of detecting at least one individual of a particular species, m is the total number of individuals of that species in all annual samples, and k is an index of temporal aggregration that is estimated from the data. For each species, we used one of two methods to estimate 'k'. When a particular species was encountered in less than '4 of the years, 'k' was obtained by solving

k *log(!+ xjk) = log(N/ j 0 ), (3) where N is the number of samples (i.e. years), f0 is the number of years in which the species wasn't detected, and xis the mean number of individuals in a sample (Bliss and Fisher 1953).

In other cases 'k' was obtained by solving the simpler form

k = x 2/(s2 -x), (4) where x is the mean number of individuals encountered each year and s2 is the variance among years of the number of individuals caught each year (Bliss and Fisher 1953).

We classified each species in each lake into groups according to their probability of being detected: 0-1.0 (all species), 0-0.25 (rare species), 0.25-0.75 (intermediate species),

0.75-1.0 (common species), and 0.25-1.0 (rare species removed). We also had a category where episodic species (those that only appeared in a single year) were removed. Based on the probability of detecting each species, we developed a presence/absence matrix that indicated whether a species was present or absent each year for the duration of the database. 97 We then used the turnover estimate (equation 1) to calculate the turnover that could be attributable to missing species that were present. This was repeated one million times to obtain the mean and distribution of sampling turnover (Arnott et al. 1998b).

Similarity indices

We used two similarity indices, Jaccard's similarity coefficient and Morisita's index of similarity, as alternative measures of community change and to determine if species turnover resulted in communities that were undergoing directional change. If community similarity decreased as the interval between measures increased, we interpreted this as directional changes in community composition as opposed to random change. Jaccard's similarity coefficient is based on presence/absence data, whereas Morisita's index is based on counts of individuals and therefore considers changes in relative abundance. Jaccard's similarity coefficient was calculated as

J = aj(a +h+c) (5) where a= the number of species present in years 1 and 2, b = the number of species present in year 1, but not 2, and c = the number of species present in year 2, but not year I. Possible values range from 0 to 1, with 0 indicating that the communities have entirely different species assemblages and 1 indicating that the communities have the same species composition. 98

Morisita's index was calculated as

(6) where c~. = Morisita's index of similarity between samples j and k, Xu. Xik = the number of individuals of species i in sample j and sample k, Ni = :EXu =the total number of individuals in sample j, Nk = :EXik =the total number of individuals in sample k, A.1 = :E[Xu(X;rl)]/Nj(Nj-

To determine whether communities changed over time, we investigated the relationship between the interval between sampling and Jaccard's similarity coefficient or

Morisita's index of similarity. A significant correlation between the time interval between censuses and similarity index indicated a directional change in species composition through time. Mantel's test was used to test for significance (Sokall979, Jackson and Somers 1989,

Arnott et al. 1998).

Species Persistence

We compared persistence of phytoplankton, zooplankton, and fishes across years by

determining the number of species that occurred in each of four persistence categories: 0-25

(rare), 26-50,51-75, and 76-100% (common) of the years.

We determined whether species persistence (the proportion of years that a species

was present) or the number of transitions from present to absent and vice versa was related to 99 abundance. Taxa with low abundance are more likely to go extinct than those with high abundance because of vagaries in birth and death rates (Pimm eta!. 1988). In addition, several studies have indicated that variability in abundance over time (measured as the coefficient of variation) is related to the number of transitions between presence and absence

(Pimm eta!. 1988, Schoener and Spiller 1992). We used linear regression to determine whether persistence was related to log density or the coefficient of variation of density.

Linear regression was also used to investigate the relationship between the number of transitions between presence and absence for each species and its coefficient of variation of density across time.

Results

Richness patterns

Annual species richness ranged from 71 to 86 (mean = 78) species for phytoplankton,

II to 17 (mean= 15) for zooplankton, and 8 to 22 (mean= 17) for fishes among lakes. In general, annual richness varied little across years. The coefficient of variation for species richness across years was low; 11 % ± 3 % for phytoplankton, 13 % ± 4 % for zooplankton, and 12% ± 6% for fishes. For most lakes observed during the study period, richness did not increase or decrease through time. However, we detected a directional change in species richness in three cases. A decline was detected in phytoplankton richness in Lake 239

(Kendal Tau, P = 0.047), an increase in zooplankton richness was observed in Crosson lake

(Kendal Tau, P = 0.0 13), and an increase in fish richness occurred in Sparkling Lake (Kendal

Tau, P = 0.009). 100 The annual richness in each lake, for all groups ranged from approximately 1/3 to 2/3 the total species pool (long-term richness) (Table 3, Fig. 2). On average, a richness estimate based on one year of sampling accounted for only half of the total species pool observed during the study period. For phytoplankton and zooplankton, asymptotic richness was frequently higher than the cumulative richness, suggesting that even 10 years of sampling is not enough to detect the entire species pool (Table 3). After a decade of sampling, only 92% of the phytoplankton, 85% of the zooplankton, and 90% of the fishes were detected. The asymptotic and cumulative richness for fishes in the NTL-LTER was similar for most lakes, suggesting that in most cases, 16 years of sampling is adequate to detect the entire species pool (on average, 97% of the species pool was detected).

Turnover Rates

Apparent turnover rates were high and similar for each of the taxonomic groups; approximately 15-20% of the species changed per year (Table 4). Although turnover rates were generally high, they varied considerably from year to year within each lake. Among year variation in turnover rates within lakes was high for fishes and zooplankton (for fishes, average CV of turnover among years = 0.40, range= 0.28-0.51; for zooplankton average CV

= 0.44, range= 0.26-0.76). In contrast, phytoplankton apparent turnover rates tended to be similar from year to year (average CV in turnover among years= 0.20, range= 0.15-0.24).

Within each taxonomic grouping, apparent turnover rates varied among lakes (ANOV A main effect lake, P < 0.001 for fishes, P <0.05 for zooplankton, and P < 0.001 for phytoplankton).

For fishes, apparent turnover rate was highest in Crystal Lake (Tukey, P < 0.001) (Table 4). lOt Zooplankton apparent turnover rate was highest in Heney Lake (Tukey, P < 0.05) and phytoplankton apparent turnover rates were significantly different between L224 and L239

(Tukey, P < 0.001), with L224 having the highest mean apparent turnover rate.

Monte Carlo simulations indicated that sampling turnover was high for all three groups of organisms. Sampling error was highest for fishes, exceeding the apparent turnover in each of the 5 lakes. For zooplankton in 6 of eight lakes, sampling turnover was as high or higher than apparent turnover (Arnott eta!. 1998b ). Phytoplankton in each of the ELA lakes had the lowest sampling turnover, relative to apparent turnover.

Both apparent and sampling turnover rates were influenced by the probability of detecting species. For all taxonomic groups, rare species (0- 0.25 probability of detection) had the highest apparent and sampling turnover rates, followed by intermediate species (0.25-

0.75 probability of detection) (Fig. 3-5). Common species (0.75-1.0 probability of detection) had the lowest apparent and sampling turnover rates because they were generally detected each year. In all lakes, apparent turnover was largely accounted for by sampling turnover for the rare species. This was also true for the intermediate species, except for zooplankton in

Chub and Red Chalk Lakes. Although apparent turnover was higher than sampling turnover for phytoplankton in all lakes, the relative magnitude was similar suggesting that even for

common species, sampling error may be driving many of the observed changes in species

composition.

Because sampling error was highest in the rare species, we removed them from the

analyses in an attempt to lower the overall sampling turnover. We defined rare in two ways;

1) episodic species that were present in only one year throughout the time record, and 2) 1111

!02 species with a probability of detection lower than 0.25. Removal of rare species, however, did not lower sampling turnover relative to apparent turnover (Fig. 3-5). Both apparent and sampling turnover decreased by approximately the same magnitude, so that even with rare species removed much of the apparent turnover could be explained by sampling error.

Similarity indices

The communities in many of the lakes became more dissimilar as the time interval between samples increased. This was true for similarity based on presence/absence, which is highly sensitive to rare species and for similarity based on abundance, which is more

sensitive to shifts in dominance (Table 5). These changes in community composition were

most striking for phytoplankton. Phytoplankton in the 6 ELA lakes exhibited an increased

dissimilarity over time except in L373 when using the abundance-based index. The results of

the two similarity indices for fish communities in NTL-LTER lakes were in strong

agreement. Fish community relative abundance in each of the five lakes became increasingly

dissimilar over time. Likewise, we detected a significant decrease in presence/absence based

similarity for all lakes except Trout Lake. For zooplankton, the two indices were not

consistently similar. Both the abundance based and the presence/absence based indices

suggested that zooplankton communities in Blue Chalk and Chub Lakes underwent a

directional change in community composition. The composition of Dickie, Harp, and Plastic

Lakes, however, changed only when considering presence/absence and the species

composition in Red Chalk changed only when considering abundance based similarity. r

103

Species Persistence

In general, communities of all three groups exhibited a bimodal distribution of species across time (Fig. 6). Species were either rare or common, with few having an intermediate persistence. On average, almost a third of the phytoplankton species were present less than 25% of the years. Approximately 50% of the phytoplankton species were detected more than 75% of the time, with an average of 20% being detected every year. The persistence pattern for zooplankton was similar; approximately one third of the species were rare, and almost 50% were common (detected >75% of years). However, the proportion of species that were detected every year was higher for zooplankton (39%) than for phytoplankton (20%) and for fishes (21 %). The proportion of rare species (detected in less than 25% of the years) was high for fishes, although much of this is because Crystal Lake

had a high frequency ofrare taxa (68%). On average, 39% of the fishes were rare, whereas

approximately a third were present in greater than 75% of the years. Only 21% were

detected every year.

Persistence verses density

Species with high density tended to be detected more frequently than species with

low mean density. Log mean annual density and persistence were positively correlated for

fish, zooplankton, and phytoplankton (P

phytoplankton rare and abundant species, however, was small (mean persistence for lower

density quartile=0.42, mean persistence for upper density quartile=0.73). In contrast, 104 differences in persistence between rare and abundant taxa for zooplankton and fishes was much greater (mean persistence for lower density quartile=0.20 for zooplankton, 0.12 for fishes, mean persistence for upper density quartile=0.94 for zooplankton, 0.89 for fishes).

The coefficient of variation (CV) of density was a better predictor of persistence than was log density for phytoplankton and fishes, but not for zooplankton. The number of transitions for fishes was not significantly related to the coefficient of variation or zooplankton (except

Crosson Lake, P=0.02). For ELA phytoplankton, the relationship was significant (P<0.05) for all lakes except L224.

Discussion

Richness estimates

The size of the total species pool (long-term richness) of a lake could not be predicted very

precisely from a single year for phytoplankton or zooplankton (Spearman Rank Correlation;

phytoplankton r, = 0.54, P>0.05, zooplankton r, = 0.00, P>0.05). Short-term richness tended

to drastically underestimate long-term richness in most cases (Fig. 2). The richness ranking

among lakes for phytoplankton and zooplankton was dependent on the time scale of the

study. This discrepancy between short-term and long-term richness for a lake is related to

the species turnover rate among years. High apparent turnover rates result in larger

differences between short-term and long-term richness. For fishes, in contrast with

phytoplankton and zooplankton, the richness ranking of lakes based on mean annual richness

was the same as the rank order for the long-term asymptotic richness. Short-term richness,

however, was approximately half of the long-term richness estimate. Interestingly, annual 1

105 richness estimates do not appear to be influenced by species turnover rates for any of the groups of organisms. Annual richness tended to remain consistent from year to year. That is, although turnover rates were high, we usually detected the same number of species every year. We did, however, detect a directional change in species richness through time in three communities; phytoplankton in L239, zooplankton in Crosson Lake, and fishes in Sparkling

Lake. Species richness has been shown to be related to lake surface area in several previous

studies (Barbour and Brown 1974, Browne 1981, Dodson 1992; but see Patalas and Saiki

1993). This relationship was not common in our analyses, however. We detected a species­

2 area relationship only for fishes in NTL-LTER lakes (linear regression, P=0.02, R =0.90,

Fig. 7). Fish sampling included both pelagic and littoral samples. As the lake area increased,

the number of different littoral habitat types also increased. Benson and Magnuson ( 1992)

found that fish diversity was related to habitat heterogeneity, especially the diversity of depth

gradients. Larger lakes offer greater habitat heterogeneity and therefore, greater

heterogeneity in species composition.

In contrast to our results for fishes, phytoplankton and zooplankton richness were not

related to lake surface area (P=0.84 for phytoplankton and P=0.53 for zooplankton), possibly

because the range in lake size was small (Table 2). Another possible explanation is that, in

contrast to sampling methodologies for fishes, we only sampled the pelagic region for

zooplankton and phytoplankton. For lakes in the size range of our study lakes, the pelagic

region may be relatively uniform with regard to horizontal habitat patchiness (Patalas and 106 Saiki 1993). Therefore, increasing lake surface area may not increase the number of habitats available to zooplankton and phytoplankton.

Turnover rates

Variation in turnover rates among lakes increased from phytoplankton to zooplankton to fishes. Apparent phytoplankton turnover rates were fairly similar, ranging from 16 to 21% per year. Zooplankton and fish apparent turnover ranged somewhat more broadly from 11 to

28% to 14 to 36% per year, respectively. We are uncertain whether this variation reflects

differences among the groups of organisms or among lake districts. For example, NTL­

LTER lakes span a wide range of surface areas and depths (Table 2) whereas ELA and

Dorset lakes are more similar in size and depth. The low variation among phytoplankton

turnover estimates may result from low variation in the physical and chemical characteristics

of the lakes.

Using three independent measures of community change, apparent turnover rates and

directional change in community similarity, it appeared that aquatic communities, in

relatively undisturbed reference sites, changed in composition through time. High apparent

turnover rates and Jaccard's similarity coefficient suggest that species replacements are

frequent. But, many of these observed changes may result from our failure to reliably detect

species in each year they occur. High sampling turnover in each of the groups of organisms

indicates that this is the case. Presence/absence-based indices tend to be highly influenced

by the presence of rare species that flicker above and below our level of detection. These

species probably have a large influence on our measure of community change, but may have !07 little immediate influence on ecological processes occurring in the lake. Furthermore, because sampling error associated with rare species is high, we cannot determine if observed changes in composition are the result of immigration and extinction events or low detection probabilities. Changes in similarity based on abundance provide evidence that community structure is dynamic and may indicate important ecological shifts through time, but they do not necessarily provide evidence of species shifts (Cisneros 1993).

Because some species are missed in routine sampling programs, changes in community composition may result from sampling error, in addition to immigration and extinction of species. Our Monte Carlo simulations suggested that uncertainty associated

with sampling error is high for each taxonomic group. In fact, sampling error could account

for as much as 60-80% of the observed phytoplankton turnover, 80-100% of the observed

zooplankton turnover, and 100% of the fish turnover. This does not necessarily mean that

actual immigration and extinction rates are low. Rather, we don't have confidence in our

ability to reliably detect some species and therefore, do not know if the community is

changing because of gains and losses of species or because we occasionally do not detect rare

species. This is particularly true in the fish communities where sampling error was higher

than apparent turnover in each lake. We know from intensive population studies of fishes in

the LTER lakes, however, that there have been at least two invasions into Crystal Lake and

likely one extinction in Sparkling Lake (Cisneros 1993, Hrabik eta!. 1998). Using an

aggregate measure, such as turnover, we were unable to distinguish those invasions and

extinctions from the many year! y gains and losses of rare species. 108 We expected to detect differences in turnover rates among groups of organisms that were related to life history traits. Fish, zooplankton, and phytoplankton have different life

history traits (Table 6) and therefore, different probabilities of immigration, extinction, and

turnover rates. Characteristics associated with rapid response to environmental change result

in high turnover rates and are strongest for phytoplankton and weakest for fishes. Short

generation times, asexual reproduction, and the potential for high population growth rates

enable new phytoplankton immigrants to become dominant members of the phytoplankton

community in a relatively short period of time if environmental conditions are favorable.

Although some species of fishes have high fecundity and are capable of producing large year

classes, their probability of success is likely lower than zooplankton and phytoplankton. In ' ' ! addition, high dispersal rates of phytoplankton (Round 1971) and zooplankton (Schlichting

and Sides 1969, Maguire 1963, Jenkins 1997) suggest that phytoplankton and zooplankton

turnover would be higher than fish turnover.

Apparent turnover rates were high and similar among different groups of species at

our different study sites. This is in contrast with our expectations based on life history

characteristics and previous studies that demonstrated a correlation between generation time

and turnover rate (Schoener 1983). Comparing results of several studies, Schoener (1983)

found that short-lived, fast-reproducing organisms tended to have high turnover rates (e.g.,

protozoans), whereas long-lived, slow-reproducing organisms had low turnover rates (e.g.,

birds, lizards, vascular plants). Our results were surprising, given differences among life

history traits, dispersal abilities and longevity of the three groups of organisms we studied.

This unexpected result was resolved when we considered the contribution of sampling error 109 to turnover rates. Sampling error was extremely high for all species, but particularly high for fishes. In fact, turnover that could be attributed to our failure to detect species was higher than apparent fish turnover rates based on observation.

Sampling error was an important factor contributing to apparent turnover rates in each of the three taxonomic groups. Based on the amount of water sampled, we expected phytoplankton to have the highest sampling turnover. Annual phytoplankton samples represent only 60 ml of water and zooplankton samples represent, on average 185 L of water.

It is more difficult to determine how much water was sampled for the fishes because many of the sampling methods used were passive. However, it can be reasoned that the volume of water sampled was orders of magnitude larger than for phytoplankton and zooplankton.

Because sample volume relative to lake volume was low for phytoplankton and zooplankton, we expected sampling error to be high. That is, we expected to miss species in our samples because of spatial heterogeneity of plankton populations (Harris and Smith 1977, Malone and

McQueen 1983). Despite the low sample volume for phytoplankton, sampling error was relatively low for phytoplankton (mean across lakes= 12%/year) compared with zooplankton

( 17%/year) and fishes (22%/year), although sampling error for all groups were of the same order of magnitude. One explanation is that spatial heterogeneity is a more important factor for fishes than zooplankton and phytoplankton. Horizontal heterogeneity is less than vertical heterogeneity for plankton communities (Pinel-Alloul et al. 1991) and is relatively low in small to medium sized lakes (89 to 2219 ha surface area, Patalas and Saiki 1993). Because our phytoplankton and zooplankton samples vertically integrated the water column and many of the ELA and Dorset Lakes are small to medium in size, spatial heterogeneity may be a 110 relatively minor component of sampling error. Spatial heterogeneity is certainly important for fishes, some of which exhibit schooling behaviours and diurnal foraging patterns (Benson and Magnuson 1992). Additionally, our fish sampling incorporated more habitat heterogeneity, encompassing both pelagic and littoral areas of the lakes, whereas zooplankton and phytoplankton samples were strictly pelagic. Spatial heterogeneity in fish populations may help explain high sampling error, despite our intensive sampling program.

An additional factor that may contribute to high sampling turnover is the presence of rare species. On average, fishes had a higher percentage of rare species (present less than

25% of the years) than either zooplankton or phytoplankton. Rare species influence the extent of sampling turnover because episodic species tend to have low abundance (frequently only one individual is caught) and therefore contribute greatly to sampling error in the Monte

Carlo simulations. We are uncertain what status rare fish species have in lake communities.

They may be species that have escaped anglers' bait buckets and temporarily survived, but did not reproduce in the lake or perhaps they are species that maintain low population densities and are, therefore, usually missed in the regular sampling regime.

The relative importance of population size and variability and how they interact with each other to influence apparent and sampling turnover rates is uncertain. As with Magnuson eta!. (1994), we detected a relationship between persistence and fish abundance. In addition, the CV of fish density was significantly related to persistence in the 16-year database, although this relationship wasn't significant for the 11-year subset used by Magnuson et a!.

(1994). These relationships are probably related more to sampling error than actual turnover. lll Species with low abundance have a lower probability of being detected each year and therefore their persistence across time is low. Likewise, species with high variability in their population abundance will not be detected reliably each year. Because of the potentially high

sampling error associated with fishes it is difficult to make conclusions regarding factors that are driving patterns in persistence.

Comparison with other estimates of species richness

Recently, other researchers have re-initiated investigations that calculate species

richness and turnover, taking into account the problem of failing to detect all species in a

system (Boulinier eta!. 1998, Hines et al, in press, Nichols eta!, in press). They have used a

capture-recapture approach that was initially developed for estimating population size, but

can be adapted for estimation of species richness (Burnham and Overton 1979). One of their

approaches requires sampling an animal community at two different times (or locations)

(primary samples). Within each of these time periods, a series of replicates (e.g. a series of

quadrats sampled at close time intervals or at several locations within the area of interest) is

taken (secondary samples). The frequency distribution of species detected in secondary

samples is combined with a jackknife estimator that accounts for heterogeneity among

species in their detection probabilities (Burnham and Overton 1979, Boulinier eta!. 1998)

and can be used to estimate species richness, the number of new colonists, and extinction

probability.

We compared our approach for estimating turnover rates and species richness with

the capture-recapture jackknife approach using a subset of our database where we had 1!2 sufficient replicates. We found that, at least for this small subset of data, our results were comparable. We estimated phytoplankton turnover rates between two years in L239 as 6.4% using the jackknife procedure developed by Hines eta!. (in press) and used by Nichols eta!

(in press) to estimate turnover rates. Calculations using our Monte Carlo simulation model estimate the difference between phytoplankton apparent and sampling turnover rates for

L239 to be 4.8% per year. Although this value is similar to the estimate of actual turnover obtained using the jackknife procedure, we are uncertain what it represents. Sampling error

can increase or decrease apparent turnover because undetected species can increase apparent

extinction rates when the species is previously detected or they can increase apparent

immigration rates when previously undetected species are suddenly detected.

A primary difference between our approach and the capture-recapture jackknife

procedure used by Nichols eta! (in press) is that our values represent long-term averages

between adjacent years, whereas the Nichols et a!. estimates represent a snapshot view of

turnover between two years (or sites). Previous studies with zooplankton (Arnott eta!.

!998b) and fishes (Magnuson et a!. 1994) have suggested that turnover rates vary

considerably from year to year and that single estimates between two points in time may not

reflect the long-term change in community composition. In particular, planktonic organisms

with short generation times and resting stages are probably highly responsive to annual

changes in environmental conditions, which in turn, probably vary considerably in magnitude

from one year to the next. Therefore, long-term estimates are necessary to encompass natural

variability in species immigration and extinction rates. 113 Implications

Sampling error is an important component of species turnover in aquatic ecosystems.

Much of the turnover observed in sampled systems may result from missing species that are actually present. Can we determine how much a community is changing given our current sampling techniques? Sampling and counting protocols used for these analyses are typical of those used for many research programs. In fact, these are some of the more complete databases available, considering their long duration and consistent methodology. Despite this, uncertainly associated with missing rare species was high. This was particularly important for the zooplankton and fish databases. Apparent turnover rates were high, but we could not determine how much of this annual change in species composition could be attributed to immigration and extinction of species. Our study takes a step toward developing methods to quantify species turnover in sampled communities. We have shown that sampling error, in the sense of our ability to detect rare species, in aquatic communities is high. The next step is to develop methods that account for this error in the estimation of community change. Methods developed for capture-recapture have proven useful for

spatially heterogeneous populations where replicate samples have been taken over space

(Nichols et al. in press, Boulinier et al. 1998). While spatial heterogeneity in aquatic

ecosystems is an important component of population variability (Benson and Magnuson

1992, Pinel-Alloul 1995), temporal variability (particularly in plankton) is likely a more

important factor in temperate lakes (Evans and Sell 1983). As a result plankton samples tend

to be taken as time series, with no true replicates being taken. Therefore, methods for

estimating species turnover (immigration and extinction) which incorporate temporal ll4 variation need to be developed.

Quantifying stability of communities is an important step in understanding how ecosystems respond to human-induced stresses or disturbances. We have demonstrated that despite extensive, long-term databases, we were unable to quantify actual turnover rates for aquatic organisms. Apparent turnover rates were high and similar among three groups of organisms with very different life history characteristics and dispersal probabilities. Our estimation of sampling turnover indicated that much of the apparent turnover could be attributed to our failure to reliably detect species in our samples, even though they may be present in the lake. Methods must be developed to parse out sampling error from estimates of apparent turnover. We have taken an initial step by developing a new and unique method for quantifying potential sampling error.

Acknowledgments

We thank Trevor Pawson, Robert Girard, Martyn Futter, and Paul Hanson for technical

assistance and Bill Geiling for counting the Z<)Oplankton samples. Janet Fischer provided

helpful reviews. Craig Stow, Conrad Lamond, Keith Somers, and Steve Carpenter provided

advice on analyses. The Dorothy Powers Grant and Eugene Lodewick Grant Scholarship

Fund provided financial support for S .E. A. Support for research was provided by the Ontario

Ministry of the Environment and the North Temperate Lakes Long-Term Ecological

Research project funded by the National Science Foundation, Grants BSR14330 and

DEB9011660. Support for ELA research was provided by the Department of Fisheries and

Oceans Canada. S. Kasian and M. Stainton supplied the chemical data. 115

References

Arnott, S. E., J. J. Magnuson and N.D. Yan. 1998. Crustacean zooplankton species richness: single- and multiple-year estimates. Canadian Journal of Fisheries and Aquatic Sciences 55: 1573-1582.

Arnott, S. E., N.D. Yan, J. J. Magnuson and T. Frost, M. 1998. Inter-annual variability and species turnover of crustacean zooplankton in shield lakes. Canadian Journal of Fisheries and Aquatic Sciences In press.

Barbour, C. D. and J. H. Brown. 1974. Fish species diversity in lakes. The American Naturalist 108:473-489.

Benson, B. J. and J. J. Magnuson. 1992. Spatial heterogeneity of littoral fish assemblages in lakes: relation to species diversity and habitat structure. Canadian Journal of Fisheries and Aquatic Sciences 49:1493-1500.

Bliss, C. I. and R. A. Fisher. 1953. Fitting the negative binomial distribution to biological data. Biometrics 176-200.

Boulinier, T., J.D. Nichols, J. R. Sauer, J. Hines, E. and K. H. Pollock. 1998. Estimating species richness: the importance of heterogeneity in species delectability. Ecology 79:1018-1028.

Brezonik, P. L., J. G. Eaton, T. M. Frost, P. J. Garrison, T. K. Kratz, C. E. Mach, J. H. McCormick, J. A. Perry, W. A. Rose, C. J. Sampson, B. C. L. Shelley, W. A. Swenson and K. E. Webster. 1993. Experimental acidification of Little Rock Lake, Wisconsin: Chemical and biological changes over the pH range 6.1 to 4.7. Canadian Journal of Fisheries and Aquatic Sciences 50: 1101-1121.

Brown, J. H. 1971. Mammals on mountaintops: nonequilibrium insular biogeography. American Naturalist 105:467-478.

Browne, R. A. 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. Journal of Biogeography 8:75- 83.

Burnham, K. P. and W. S. Overton. 1979. Robust estimation of population size when capture probabilities vary among animals. Ecology 60:927-936.

Cisneros, R. 0. 1993. Detection of cryptic invasions and local extinctions of fishes using a long-term database. Thesis. University of Wisconsin-Madison, Madison, Wisconsin. 116 Diamond, J. M. 1969. Avifauna! equilibria and species turnover rates on the Channel Islands of California. Proceedings of the National Academy of Science 64:57-63.

Dillon, P. J., R. A. Reid and H. E. Evans. 1993. The relative magnitude of phosphorus sources for small, oligotrophic lakes in Ontario, Canada. Verhandlungen Intemationale Vereinigung fur Theoretische und Angewandte Limnologie 25:355- 358.

Dobson, A. P., Rodriguez, J.P., Roberts, W. M., and Wilcove, D. S. 1997. Geographic distribution of endangered species in the United States. Science 275:550-553.

Dodson, S. 1992. Predicting crustacean zooplankton species richness. Limnology and Oceanography 37:848-856.

Evans, M.S. and D. W. Sell, 1983. Zooplankton sampling strategies for environmental studies. Hydrobiologia 99:215-223.

Frost, T.M., and S.R. Carpenter, and T.K. Kratz. 1992. Choosing ecological indicators: Effects of taxonomic aggregation on sensitivity to stress and natural variability. Pages 215-227 in: D.H. McKenzie, D.E. Hyatt, and V.J. McDonald, eds. Ecological Indicators (Volume 1). Elsevier Applied Science Publishers, Essex, England. 810pp.

Goodman, D. 1987. The demography of chance extinction. 11-34 11-34 in M. E. Soule, M. E. Soule. Viable Populations for Conservation. Cambridge University Press, Cambridge.

Hairston Jr., N. G. 1996. Zooplankton egg banks as biotic reservoirs in changing environments. Limnology and Oceanography 41:1087-1092.

Hansson, L.-A., L. G. Rudstam, T. B. Johnson, P. Soranno andY. Allen. 1994. Patterns in algal recruitment from sediment to water in a dimictic, eutrophic lake. Canadian Journal of Fisheries and Aquatic Sciences 51:2875-2833.

Harris, G. P. and R. E. H. Smith. 1977. Observations of small-scale patterns in phytoplankton populations. Limnology and Oceanography 22:887-899.

Reeky, R. E., P. Campbell and D. M. Rosenberg. 1994. Introduction to experimental lakes and natural processes: 25 years of observing natural ecosystems at the Experimental Lakes Area. Canadian Journal of Fisheries and Aquatic Sciences 51:2721-2722.

Hines, J. E., T. Boulinier, J.D. Nichols, J. R. Sauer and K. H. Pollock. in press. COMDYN: software to study the dynamics of animal communities using a capture-recapture approach. Bird Study 117 Hrabik, T. 1995. Feeding and Distribution of Rainbow Smelt in Wisconsin: Effects on Native Fishes. Thesis. University of Wisconsin-Madison, Madison, Wisconsin.

Hrabik, T. R., J. J. Magnuson, and A. S. McLain. 1998. Predicting the effects of rainbow smelt on native fishes in small lakes: evidence from long-term research on two lakes. Can. J. Fish. Aquat. Sci. 55:1364-1371.

Hunt, G. L. and M. W. Hunt. 1974. Trophic levels and turnover rates: the avifauna of Santa Barbara Island, California. The Condor 76:363-369.

Jackson, D. A. and K. M. Somers. 1989. Are probability estimates from the permutation model of Mantel's test stable? Canadian Journal of Zoology 67:766-769.

Jenkins, D. G. 1995. Dispersal-limited zooplankton distribution and community composition in new ponds. Hydrobiologia 313/314:15-20.

Leigh, E. G. L. 1981. The average lifetime of a population in a varying environment. Journal of Theoretical Biology 90:213-249.

Lodge, D. M. 1993. Biological invasions: Lessons for ecology. Trends in Ecology and Evolution 8:133-137.

Lodge, D. M. 1993. Species invasions and deletions: Community effects and responses to climate and habitat change. 367-387 367-387 in P.M. Kareiva, J. G. Kingsolver and R. B. Huey, P.M. Kareiva, J. G. Kingsolver and R. B. Huey. Biotic Interactions and Global Change. Sinauer Associates Inc., Sunderland, Massachusetts.

Lyons, J. and J. J. Magnuson. 1987. Effects of Walleye predation on the population dynamics of small littoral-zone fishes in a northern Wisconsin lake. Transactions of the American Fisheries Society 116:29-39.

MacArthur, R. H. and E. 0. Wilson. 1967. The Theory of Island Biogeography. Princeton University Press, Princeton.

Magnuson, J. J., C. J. Bowser and T. K. Kratz. 1984. Long-term ecological research (LTER) on north temperate lakes of the United States. Internationale Vereinigung fiir Theoretische und Angewandte Limnologie, Verhandlungen 22:533-535.

Magnuson, J. J., T. W. M., A. Banerjee, J. Toivonen, 0. Sanchez and M. Rask. 1998. Isolation versus extinction in the assembly of fishes in small northern lakes. Ecology In Press: 118 Magnuson, J. J., B. J. Benson and A. S. McLain. 1994. Insights on species richness and turnover from long-term ecological research: fishes in north temperate lakes. American Zoologist 34:437-451.

Maguire Jr., B. 1963. The passive dispersal of small aquatic organisms and their colonization of isolated bodies of water. Ecological Monographs 33: 161-185.

Malone, B. J. and D. J. McQueen. 1983. Horizontal patchiness in zooplankton populations in two Ontario kettle lakes. Hydrobiologia 99:101-124.

Margules, C. R., A. 0. Nicholls and M. B. Usher. 1994. Apparent species turnover, probability of extinction and the selection of nature reserves: a case study of the Ingleborough limestone pavements. Conservation Biology 8:398-409.

McLain, A. S. 1991. The invasion of a non-native species into pelagic fish assemblages. Thesis. University of Wisconsin-Madison, Madison.

McQueen, D. J. and N.D. Yan. 1993. Metering filtration efficiency of fresh water zooplankton hauls: reminders from the past. Journal of Plankton Research 15:57-65.

Milberg, P. and M. L. Hansson. 1993. Soil seed bank and species turnover in a limestone grassland. Journal of Vegetation Science 4:35-42.

Nauwerck, A. 1963. Die Beziehungen zwischen zooplankton und Phytoplankton en see Erken. Symb. Bot. Upsal. 17: 163 p.

Nichols, J.D., T. Boulinier, J. E. Hines, K. H. Pollock and J. R. Sauer. in press. Estimating rates of local species extinction, colonization and turnover in animal communities. Ecological Applications

Patalas, K. and A. Saiki. 1993. Spatial variation of crustacean plankton in lakes of different size. Canadian Journal of Fisheries and Aquatic Sciences 50:2626-2640.

Pimm, S. L. 1993. Biodiversity and the balance of nature. 347-359 347-359 in E.-D. Schulze and H. A. Mooney, E.-D. Schulze and H. A. Mooney. Biodiversity and Ecosystem Function. Springer-Verlag, Berlin.

Pimm, S. L., G. J. Russell, J. L. Gittleman and T. M. Brooks. 1995. The future of biodiversity. Science 269:347-350.

Pinel-Alloul, B. 1991. Spatial distribution patterns in freshwater macrozooplankton: variation with scale. Canadian Journal of Zoology 69:1557-1570. ll9 Pinel-Alloul, B., 1995. Spatial heterogeneity as a multiscale characteristic of zooplankton community. Hydrobiologia 300/301:17-42.

Richter-Dyn, N. and N. S. Goel. 1972. On the extinction of a colonizing species. Theoretical Population Biology 3:406-433:

Rogers, R. W. and D. Morrison. 1994. Floristic change on Heron Island, a coral cay in the Capricorn-Bunker Group, Great Barrier Reef. Australian Journal of Botany 42:297- 305.

Round, F. E. 1971. The growth and succession of algal populations in freshwaters. Mitteilungen der internationale Vereinigung fur theoretische und angewandte Limnologie 19:70-99.

Rudstam, L. G. and J. J. Magnuson. 1985. Predicting the vertical distribution of fish populations: analysis of cisco, Coregonus artedii, and yellow perch, Perea flavescens. Canadian Journal of Fisheries and Aquatic Sciences 42:1178-1188.

Rusch, G. and E. van der Maarel. 1992. Species turnover and seedling recmitment in limestone grasslands. Oikos 63: 139-146.

Sampson, C. J., P. L. Brezonik, T. M. Frost, K. E. Webster and T. D. Simonson. in press. Experimental acidification of Little Rock Lake, Wisconsin: The first four years of chemical and biological recovery. Water, Air, and Soil Pollution 85:1713-1719.

Sanderson, B. L. and T. M. Frost. 1996. Regulation of dinoflagellate populations: relative importance of grazing, resource limitation, and recruitment. Canadian Journal of Fisheries and Aquatic Sciences 53: 1409-1417.

Schindler, D. W., K. H. Mills, D. F. Malley, D. L. Findlay, J. A. Shearer, I. J. Davies, M.A. Turner, G. A. Linsey and D. R. Cmikshank. 1985. Long-term ecosystem stress: the effects of years of experimental acidification on a small lake. Science 228:1395-1401.

Schoener, T. W. 1983. Rate of species turnover decreases from lower to higher organisms: a review of the data. Oikos 41:372-377.

Schoener, T. W. and D. A. Spiller. 1987. High population persistence in a system with high turnover. Nature 330:474-477.

Schoener, T. W. and D. A. Spiller. 1992. Is extinction rate related to temporal variability in population size? An Empirical answer for orb spiders. The American Naturalist 139:1176-1207. 120 Simberloff, D. S. and E. 0. Wilson. 1969. Experimental zoogeography of islands. The colonization of empty islands. Ecology 50:278-296.

Sokal, R. R. 1979. Testing statistical significance of geographic variation patterns. Systematic Zoology 28:227-232.

Terborgh, J. and J. Faaborg. 1973. Turnover and ecological release in the avifauna of Mona Island, Puerto Rico. The Auk 90:759-779.

Toft, C. A. and T. W. Schoener. 1983. Abundance and diversity of orb spiders on 106 Bahamian islands: biogeography at an intermediate trophic level. Oikos 41:411-426.

Tonn, W. M., J. J. Magnuson, M. Rask and J. Toivonen. 1990. Intercontinental comparison of small-lake fish assemblages: the balance between local and regional processes. The American Naturalist 136:345-75.

Webster, K. E., P. A. Soranno, S. B. Baines, C. J. Bowser, T. K. Kratz, P. J. Dillon, P. Campbell, E. J. Fee and R. E. Hecky. In review. Structuring features of lake districts: geomorphic and landscape controls on lake chemical responses to drought. Freshwater Biology

Wright, S. J. 1985. How isolation affects rates of turnover of species on islands. Oikos 44:331-340.

Yan, N.D., W. Keller, K. M. Sommers, T. W. Pawson and R. E. Girard. 1996. The recovery of crustacean zooplankton from acid and metal contamination: comparing manipulated and reference lakes. Canadian Journal of Fisheries and Aquatic Sciences 53: 1301-1327. Table 1: Study site characteristics that are related to dispersal of organisms.

Characteristic ELA Dorset NTL-LTER

Average wind speed 10 kph 5.6 kph 10 kph

No. ice-free days 200 225 222

No. stream connections 1 3 1.6

Density of lakes 19.2% 13.2% 16%

Lake size range 5-50 ha 20-100 ha 35-1600 ha

Surrounding vegetation forested forested forested

Human activity low medium high Table 2: Morphometric and chemical data for the study lakes during the period 1978 to 1989. Values represent the mean of all

samples taken during the ice-free season (generally biweekly to monthly) over the twelve-year study period. Standard deviations for chemical data are in parentheses.

Lake Latitude Longitude Area Mean Max. Alkalinity Conductivity DOC pH TP

(N) (W) (ha) Depth(m) Depth(m) mg!L CaC03 (f.IS) (mg!L) (f.! giL)

Blue Chalk 45° 11' 78° 56' 52.4 8.5 23.0 4.1(0.3) 28.9(0.7) 1.8(0.2) 6.6(0.2) 6.3(0.9)

0 4.7(0.3) 5.6(0.1) 10.4(1.4) Chub 45 13' 780 59' 34.4 8.9 27.0 0.8(0.2) 27.9(0.6) 0 4.0(0.2) 5.6(0.1) 11.0(1.4) Crosson 45 05' 78° 02' 56.7 9.2 25.0 0.6(0.2) 26.0(0.5)

Dickie 45° 09' 79° 05' 93.6 5.0 12.0 1.2(0.4) 28.3(0.8) 5.0(0.3) 5.9(0.2) 11.4(2.1)

0 3.8(0.2) 6.3(0.1) 7.5(1.1) Harp 45 23' 78° 08' 71.4 13.3 37.5 3.3(0.2) 34.4(0.6) 0 2.8(0.3) 5.8(0.2) 7.0(1.4) Heney 45 08' 79° 06' 21.4 3.3 5.8 0.5(0.2) 25.3(0.8)

Plastic 45° 11' 78° 50' 32.1 7.9 16.3 0.5(0.2) 22.5(0.6) 2.2(0.1) 5.7(0.1) 6.0(1.0)

Red Chalk 45° 11' 78° 56' 32.1 16.7 38.0 3.3(0.2) 29.4(0.4) 2.4(0.2) 6.3(0.2) 5.1(0.7)

0 7.12 10 L222 49 70' 93° 73' 16.4 3.7 5.8 9.9 38 10.8

"' Lake Latitude Longitude Area Mean Max. Alkalinity Conductivity DOC pH TP

(N) (W) (ha) Depth(m) Depth(m) mgfL CaC03 (J.IS) (mg!L) (J.lg!L)

0 L224 49 69' 93° 72' 25.9 11.6 27.4 3.8 25 3.48 6.96 6

L239 49° 66' 93° 71' 56.1 10.5 30.4 7.8 37 5.42 7.16 7.5

0 L240 49 65' 93° 73' 44.1 6.1 13.1 6.6 34 6.24 7.17 7

0 L373 49 74' 93° 80' 28.0 10.7 21.5 8.3 33 4.2 7.55 5

0 L382 49 71' 93° 68' 5.6 5.7 9.7 4.2 28 8.4 6.96 8.5

Allequash 46° 02' 89° 37' 168.4 2.9 8.0 88 7.5 29.3

Big 46° 01' 89° 37' 396.3 7.5 21.3 49 7.3 22.5

Muskellunge

Crystal 46° 00' 89° 37' 36.7 10.4 20.4 14 6.0 8.6

0 Sparkling 46 00' 89° 42' 64.0 10.9 20.0 80 7.3 15.2

35.7 93 7.6 16.9 Trout 46° 02' 890 40' 1607.9 14.6

N w 124 Table 3: Mean annual richness is the average richness across all years. The cummulative richness is based on 10 years for the phytoplankton, 16 years for the fishes, and 10 years for zooplankton (except Heney, which was 9 years). ST (short-term):LT(long-term) ratio is the ratio between the mean annual richness and the asymptotic richness, the estimate of the size of the total species pool.

Lake Mean Annual Cumulative Asymptotic ST:LTratio

Phytoplankton

L222 78.9 144 156 0.51

L224 78.3 167 201 0.39

L239 70.9 120 142 0.50

L240 85.8 144 149 0.58

L373 75.3 131 135 0.56

L382 76.1 117 119 0.64

mean 77.6 137 150 0.53

Zooplankton

Blue Chalk 15.2 23 24 0.63

Chub 15.0 23 40 0.38

Crosson 15.2 23 24 0.63

Dickie 16.2 28 38 0.43

Harp 16.0 24 30 0.53

Heney 11.4 26 33 0.35

Plastic 12.3 20 21 0.59 125 Table 3 (continued): Richness estimates Lake Mean Annual Cumulative Asymptotic ST:LT ratio

Red Chalk 17.2 24 24 0.72 mean 14.8 24 29 0.53

Fishes

Allequash Lake 20.4 35 36 0.57

Big Muskellunge 19.2 31 31 0.62

Crystal Lake 7.8 24 27 0.29

Sparkling Lake 14.8 28 28 0.53

Trout Lake 22.4 38 38 0.59 mean 16.9 31 32 0.52 Table 4: Turnover rates for phytoplankton, zooplankton, and fishes.

Phytoplankton Apparent Sampling Zooplankton Apparent Sampling Fishes Apparent Sampling

L222 19.7 13.7 Blue Chalk 11.4 13.1 Allequash 13.6 16

L224 21.0 17.7 Chub 15.8 16.9 Big Muskellunge 13.9 19.3

L239 15.9 11.1 Crosson 12.6 13.6 Crystal 35.7 37.0

L240 16.6 10.5 Dickie 19.6 22.4 Sparkling 17.5 20.6

L373 16.5 10.0 Harp 11.4 14.9 Trout 17.7 19.0

L382 15.6 9.3 Heney 27.8 22.9

Plastic 15.7 17.3

Red Chalk 12.7 11.8

mean 17.6 12.0 15.9 16.6 19.7 22.4

'"'"' 127 Table 5: Summary statistics from Mantel's test for Jaccard's Similarity versus the number of years between sampling date. Years between sampling ranged from 1 ton. In lake with P- values less than 0.05 (boldface) the species composition became more dissimilar as the interval between sampling increased. Jaccard's Similarity Morisita's Similarity

Lake n Pearson's Significance Pearson's Significance

Coefficient Level Coefficient Level

Phytoplankton

L222 12 0.59 <0.001 0.43 0.001

L224 10 0.64 <0.001 0.60 <0.001

L239 14 0.74 <0.001 0.43 <0.001

L240 10 0.82 <0.001 0.35 0.01

L373 10 0.62 <0.001 0.04 0.42

L382 9 0.40 0.01 0.52 0.001

Zooplankton

Blue Chalk II 0.25 0.04 0.38 <0.01

Chub II 0.40 <0.01 0.42 0.001

Crosson II 0.004 0.35 -0.16 0.90

Dickie II 0.35 0.01 0.22 0.06

Harp II 0.26 0.03 0.21 0.08

Heney 8 -0.07 0.64 0.08 0.32

Plastic 9 0.41 0.01 -0.22 0.94

Red Chalk II 0.16 0.13 0.38 0.01 128 Table 5 (continued): Summary statistics from Mantel's test

Jaccard's Similarity Morisita's Similarity

Lake n Pearson's Significance Pearson's Significance

Coefficient Level Coefficient Level

Fishes

Allequash 15 0.18 0.06 0.35 0.004

Big Muskellunge 15 0.41 <0.01 0.23 0.02

Crystal Lake 15 0.18 0.05 0.54 <0.001

Sparkling Lake 15 0.44 <0.001 0.48 <0.001

Trout Lake 15 0.14 0.10 0.27 0.01 Table 6: Life history characteristics of phytoplankton, zooplankton and fishes. Life history characteristics, unless noted otherwise, were from Reynolds 1984 for phytoplankton, Dodson and Frey 1991, Williamson 1991 for zooplankton, and Becker

1983 for fishes. Trait phytoplankton zooplankton fishes

generation time hours to days 1 to several weeks 1 to 7 years

lifespan 2 to 3 days 1 to several months 2-20 years

fecundity doubling time: hours to days several 100 over lifetime lO's to 1000's

reproduction sexual, asexual, resting stages sexual, asexual, resting stages sexual

body size 4 to lOOOJ.lm 0.2 to 18.0 mm several em to> 32 em •

Dispersal high-aerial, animals, water b medium-aerial, animals, water ' low-water connections, humans 1 Patchiness cmtomd < 33 m of shoreline

5 6 1 population densities 1.5 * 10 cells/L g 0.34 ind./L h smelt: 4.7 * 10· adults/L perch: 3.1 * 10"6 adults/L

pop. variability CV = 196.1% (121.2 s.d.) i CV = 72.9% (62.9 S.d.) k CV = 220.5% (118.3 S.d.) I

1 1 pop. growth rates, r 0.2 to 2.9 day" m -0.3 day" n a: from size categories in Magnuson and Lathrop 1992

tc: b: Schlichting 1969, Schlichting and Sides 1969, Procter 1959

c: Procter 1959, Maguire 1969?

d: Harris and Smith 1977

e: Tessier, Pinel Alloul, Malone and McQueen

f: Benson and Magnuson 1992

g: mean density of all species in ELA phytoplankton database (all lakes and all years)

h: mean density of all species in Dorset zooplankton database (all lakes and all years)

I: based on accoustic estimates for fish >75 mm in Crystal Lake from 1981 to 1995.

J: Coefficient of variation (stdev/mean) across years for all phytoplankton in ELA database

k: Coefficient of variation (stdev/mean) across years for all zooplankton in Dorset database

1: Coefficient of variation (stdev/mean) across years for all fish species in LTER lakes

m: maximum exponential growth rates for freshwater phytoplankton in culture, based on Table 16 in Reynolds 1984

n: for Daphnia pulex, Riessen and Sprules 1990

w 0 131 Figure Captions

Figure l. Map showing location of the three study sites.

Figure 2. Asymptotic richness versus mean annual richness for ELA phytoplankton (open

squares), Dorset zooplankton (open circles), and NTL-LTER fishes (dark triangles). The

dashed line indicates where asymptotic richness equals mean annual richness.

Figure 3. Apparent turnover (dark bars) and sampling turnover (white bars) for

phytoplankton in each of the ELA lakes. 'all' represents turnover rates when all species

were included in the analyses. '-single' represents the turnover rates when species that

appeared in only one year were removed from the analyses. Rare represents turnover

rates of species with 0-0.25 probability of detection, intermediate is for species with 0.26-

0.75 probability of detection, and common is for species with 0.76-1.0 probability of

detection. '-rare' is for species with 0.26-1.0 probability of detection. Stars indicate that

apparent turnover is greater than 90% of the sampling turnover estimates.

Figure 4. Same as figure 3, but for Dorset zooplankton.

Figure 5. Same as figure 3, but for NTL-LTER fishes.

Figure 6. Distribution of species across years. Bars indicate the percent species in each of

four persistence categories, averaged across lakes in each database. Error bars represent 132 the standard deviation of the mean. The dark portion of the 7 6-100% bar indicates the

percent species that were detected every year.

Figure 7. The relationship between asymptotic richness and lake area for phytoplankton

(open squares), zooplankton (open circles), and fishes (dark triangles). 133

,, '

#

" { ;: ,!2' ,, "" £ j lL ~ \ ~ , " ~ ~ \~,, ""'·~~~\.• :;,• ' ~...,: \.,! ·"\ ' }_' ' " \ ~... ll "" ~ ' ,,"

a: w 1- ~ --' --' 0 w --' 1- z !34

D 200 phytoplankton ~ Do 150 Lb D •• 100 •• •• •••• •• •• zooplankton .•. · •"\ \\t'.e 50 \ ····· \'. .. ······h oJlJ... ~·~. its •• 0~·~··--~--~~---r----~---, 0 20 40 60 80 100 Mean Annual Richness

Fig. 2 135

Phytoplankton

100 Apparent L222 L224 80 n1m~ Sampling 60 t/ turnover 40 20 * * * * * * * * 0

100 L239 L240

:;g0 80 ~ Q) 'lii 60 a:... ~ 40 0 !:... ::s 20 1- * * * * * * * * 0 ,11-,,h

100 L373 L382 80

60

40 20 * * * * * * * * 0 all -single rare int com -rare all -single rare int com -rare Fig. 3 136

Zooplankton

100 Blue Chalk Chub 80 Apparent turnover \ 60 Sampling * j( turnover 40 20 0

100 Crosson Dickie 80 60 ,--, <;'?. 40 ~ ~ 20 ~ ll I{] ll ~- 0

~ i>'"' 100 0 Harp Heney = 80 '"' E-<= 60 40 20 * 0 100 Plastic Red Chalk 80 * 60 40 20 0 all -single rare int com -rare all -single rare int com -rare

Fig. 4 137

Fishes

Big Muskellunge 10 Apparent turnover Allequash 80 -::.. Sampling tumov r 60 ,I 40 20 0

Mendota ~ Crystal -;!!_ 0 100 ~ Q) 80 Ill a:- ..... 60 Q) > 40 0 .....c: 20 ::I * 1- 0

Sparkling Trout 10 80 60 40 20 ...,,.-J 0 all -single rare int com -rare all -single rare int com -rare

Fig. 5 ..-. -;!!_ 0 Phytoplankton Ill -Q) '(j 40 Q) 0. (/) 20 r::: -Q) ....0 Q) 0 a.

..-. -;!!_ 60 Zooplankton 0 -Ill .!!! 0 Q) 40 0. (/) r::: 20 -Q) ....0 Q) a. 0 ..-. -;!!_ 0 Fishes Ill -Q) '(j Q) 0. (/) r::: -Q) ....0 a.Q) 0 0-25 25-50 50-75 76-100 Persistence across Years(%)

Fig. 6 139

250 (/) (/) Q) c 200 0 phytoplankton .c (.) / ·-a: 150 0 (.) OC!:J ·- 0 0 -c. 100 -E zooplankton fish >-(/) 50 ~. <( \;;,,;? .. .. A 0 1 10 100 1000 10000 Lake Area (ha)

Fig. 7 140

CHAPTER FOUR

LAKES AS ISLANDS: BIODIVERSITY, INVASION, AND EXTINCTION4

Introduction

Lakes are like islands (Fig. 1). They tend to be isolated and can therefore, be thought of as islands of water surrounded by large areas of inhospitable land (Barbour and Brown

1974, Keddy 1976, Magnuson 1976). Because of the island-like nature of lakes, theories that have been developed for islands may also apply to lakes (MacArthur and Wilson 1967,

Gilpin and Hanski 1991, Gotelli and Kelley 1993). These theories emphasize the importance of regional population processes, suggesting that the number of species occupying a habitat is a function of processes that control immigration and extinction of local populations. Based on these theories, we expect species richness in lake-islands to be dynamic in nature. That is, the number of species present at any given time is a function of the frequency of immigration of new species and the extinction rate of species previously established in the lake.

Immigration depends on factors controlling isolation, such as the distance to the nearest source of colonists or the presence of stream connections, as well as, the intensity of interactions with established species. Isolation influences the distribution of species by defining the degree of movement of individuals to and from lake-islands. It is a function of

4 Arnott, S. E., J. J. Magnuson, S. I. Dodson. Prepared for N1L-LTER publication: Lakes in the Landscape: Long-Term Ecological Research of North Temperate Lakes. Edited by J. J. Magnuson and T. K. Kratz. 141 both the physical nature of the landscape and the vagility of the taxa under consideration.

Fish dispersal typically will be limited by stream connections and human-aided introductions

(Magnuson 1976). Plankton dispersal is less constrained because wind, birds and mammals can serve as vectors (Proctor 1965, Maguire 1969, Jenkins 1995). Therefore, a particular mode of transport between lakes may be an important factor in population dynamics for one species, but not for another.

Extinction probabilities on both islands and in lakes are influenced by a particular system's characteristics. Lake size may be important in determining species richness because it is an indication of habitat heterogeneity (Tonn and Magnuson 1982, Eadie and

Keast 1984). Larger lakes tend to have a greater diversity of habitats and can therefore,

provide suitable environments for a larger number of species. Population sizes are expected

to vary with lake size. Like islands, many lakes are small. This results in fewer resources

and leads to small population size. Small populations are susceptible to extinction, which

reduces the number of species in a habitat. Other local scale characteristics such as resource

levels, habitat complexity, and abiotic and biotic conditions also influence species

composition. Ultimately, biotic structure is controlled by both regional scale mechanisms

that determine which species can arrive at a particular lake, and local scale mechanisms that

influence the survival of species once they are in the lake. Therefore, in our examination of

factors that control species richness and invasion and extinction processes, we must broaden

our scale and consider both local and regional drivers.

In our studies of north temperate lakes, we expected to see a relationship between

extinction factors such as lake size and the number of species present. We also expected that 142 isolation factors such as distance to other lakes would be important in determining species composition. And, we expected lakes to be dynamic in their species composition with species appearing and disappearing through time with an equilibrium number of species generated by the interplay between immigration and extinction rates. In some cases, this island-biogeography framework has proven to be a useful conceptual tool. In other cases, however, we find that the systems we study are more complex and difficult to evaluate.

Specifically, we have addressed several key issues:

• Does species composition fluctuate?

• How many species are present in a lake?

• What factors influence the species richness of lakes?

• How well can we evaluate if an invasion or extinction has occurred?

• What is the relative importance of invasion versus extinction to the assemblage and

richness?

• A related issue concerning the influence of exotic species invasions on lake ecology is

treated in Chapter 8 as an external driver on lake dynamics.

II. The dynamic nature of aquatic communities

The availability of long-term databases for aquatic organisms provided us with a unique opportunity to examine the stability of species composition in lakes. We have asked,

"Do lakes house stable assemblages of species or is there a continuous turnover of species through time?'' To address this question, we examined temporal changes in community 143 composition (i.e., species turnover) for aquatic organisms in several lakes (Magnuson and

Lathrop 1992, Magnuson eta!. 1994, Arnott eta!. in review, Arnott eta!. ms).

Species turnover rates for Lake Mendota, a highly managed system, were calculated from historical fish records and recent fish surveys by the Wisconsin Department of Natural

Resources. From the 1900's to 1989, we estimated that the fish species turnover rate was approximately 2 fish species per decade (Magnuson and Lathrop 1992). Much of this species turnover, however, occurred during two periods of intense human impact (see chapter II).

During the 1920's there was a decline in small-bodied littoral fishes that probably resulted from a combination of habitat degradation associated with increased eutrophication and widespread use of herbicides and weed cutters to control macrophyte beds. Additional changes in the fish community occurred during this time resulting from intentional fish stocking and from fish being released as part of the Mississippi River Fish Rescue Program.

During the 1970's and 1980's the decline of small-bodied fishes probably resulted from increased piscivory. Stocking rates of piscivorous sport fish during this time were the highest they had been in the history of the lake. These historical records for Lake Mendota indicate that species richness and community composition has been changed dramatically during the past century. Albeit human-induced, compositional change, as in Lake Mendota, appears to be the norm rather than the exception.

Species turnover rates were also high for lakes with relatively low human impact.

We calculated species turnover rates for fishes in the Northern Highland Lake District in

Wiscsonsin (Magnuson et a!. 1994 ), zooplankton in the Dorset region in southern Ontario

(Arnott et a!. in review), and phytoplankton in the Experimental Lakes Area in northwestern 144 Ontario (Amott eta!. ms). All three regions are primarily recreational and have experienced relatively little hnman impact. The databases are suited for addressing issues of community stability because they have been methodologically consistent for over a decade. Apparent turnover rate (T) between two sampling periods was calculated as

l+E T=lOO* *r' S, +S2

where I is the number of species gained, E is the number of species lost, S 1 and Sz are the number of species present in each sample, and t is the time interval between samples

(Diamond 1969). Apparent turnover was high; on average 22% of the fishes, 16% of the zooplankton, and 18% of the phytoplankton, appeared or disappeared each year. Much of this turnover, however, can not be distinguished from sampling error. That is, despite

extensive sampling, rare species were sometimes missed even though they were present in

the lake.

To adjust for sampling error, we calculated conservative estimates of turnover rates

using two different approaches. In our initial analyses for fishes, we removed rare species

from the analysis if fewer than two individuals were encountered during the study period,

and we subtracted the turnover between adjacent years from our estimate (Fig. 2, Magnuson

eta!. 1994). Our rationale was that turnover between adjacent years probably resulted from

missing rare species and that the time interval was too short for changes resulting from

extinction and invasion. When the above adjustments were made, fish species turnover

averaged 0.44% per year for all lakes, that is, over the course of a decade one to two species

would either immigrate or go extinct (Magnuson eta!. 1994). 145 Because zooplankton and phytoplankton have shorter life cycles than fishes, changes in community composition resulting from invasion and extinction seem more likely on yearly time scales. To account for turnover attributable to sampling error we developed a Monte

Carlo simulation model where the probability of detecting each species was calculated based on the average number of individuals caught each year. Based on these probabilities, we determined the amount of turnover that could potentially result from sampling error (Fig. 3).

When we subtracted potential sampling error from our apparent turnover estimates, we could reliably detect zooplankton species turnover in only two of the eight Dorset study lakes.

Zooplankton species turnover, corrected for sampling error, averaged 0.7% per year. We could detect phytoplankton species turnover for all six ELA lakes. Phytoplankton species turnover, corrected for sampling error averaged 6.2% per year. We performed similar calculations for NTL-LTER fishes but could not reliably detect actual turnover for any of the six lakes (although, recall that Magnuson et al. (1994) calculated rates of approximately

0.44% per year). We suspect that our estimates of species turnover are underestimates. Even so, we were able to demonstrate that even for isolated, island-like habitats with relatively little human impact, movement of organisms among lakes and local extinction may be important in determining local community structure.

III. How many species are there?

The high apparent turnover of species, whether attributed to actual gains and losses of

species or sampling error, has important implications for the assessment of species richness

in aquatic habitats. For fishes, crustacean zooplankton, and phytoplankton the estimated 146 long-term species pool (based on an asymptotic estimate) was, on average, two times greater than the mean annual richness (Fig. 4, Arnott et al. ms). Richness depended on the number of years sampled (Magnuson et al. 1994, Arnott et al. 1998) and the type of index used to calculate richness (Arnott et al. 1998). Because zooplankton in each lake had different turnover rates, the rank order species richness based on short -term assessments was not correlated with the rank order richness based on long-term estimates. Much of this

discrepancy was the result of episodic species that tended to appear in lakes for on1 y one

year. For example, Heney Lake had the fewest species when considering the mean annual

richness. However, because of its high apparent turnover rate (approximately 25% per year),

it had the highest long-term, cumulative species richness.

The discrepancy between short-term and long-term richness influences the

relationship between richness and explanatory variables such as lake area, lake depth, pH,

and nutrients. The importance of each variable depends on whether the richness estimate is a

mean annual estimate or a long-term cumulative estimate (Magnuson et al. 1994, Arnott et al.

1998). For example, mean lake depth was the most important variable predicting mean

annual zooplankton species richness, explaining 61% of the variation for the Dorset lakes. In

contrast, total phosphorus concentration was the most important variable predicting long­

term cumulative species richness, explaining 45% of the variation.

These results have important implications for the assessment of species richness. The

discrepancy between short-term and long-term richness rankings may be resolved by

considering that each richness estimate may represent a different property of the lake

community. Short-term richness is a measure of the number of potentially interacting 147 species, whereas long-term richness is a measure of the potential species pool and may be related with the ability of a system to respond to stresses. These results suggest that richness estimates should be considered an index of richness for habitats where equal effort is applied to obtain the estimate. That is, because the number of species detected is dependent on the extent of the sampling approach and the method of calculating richness, caution must be exercised when searching for environmental correlates in multi-lake comparisons. At the very least, richness estimates for comparison should be based on the same intensity of sampling (i.e., number of samples within a year, number of years sampled) for all lakes in the assessment. Finally, we must also recognize that short-term richness estimates greatly underestimate the total species pool. The extent of this underestimation depends on species turnover rates within each lake.

IV. What determines the number of species in a lake?

The equilibrium theory of island biogeography suggests that the number of species present in a lake island is a function of immigration and extinction rates. What factors, however, are important in determining these rates and in setting the equilibrium number of

species? Why do some lakes have more species than other lakes? We have examined

several possible factors that regulate species richness of aquatic organisms. Although our

emphasis in these studies has been primarily on local drivers, we acknowledge the

importance of regional influences on immigration rates.

Numerous studies have indicated that species richness depends on the size of the lake.

As lake surface area increases, there is an increasing number of species in several taxa, 148 including fishes, zooplankton, macrophytes, mussels, and snails (Barbour and Brown 1974,

Lassen 1975, Browne 1981, Tonn and Magnuson 1982, Rahell986, Eadie et al. 1986, Minns

1989, Dodson 1992, Magnuson et al. 1994, Dodson et al. Ms). The relationship between species richness and lake size, however, is often influenced by other factors. The slope of the species-area curves differed depending on the composition of fish assemblages (Tonn and

Magnuson 1982), the alkalinity of lakes in the dataset (Rahel 1986), and the method of

estimating richness (Magnuson et al. 1994). Our studies indicate that, in ger,eral, lake size

(or habitat heterogeneity) is an important determinant of the number of species able to

coexist in a lake. We caution, however, against broad scale comparisons of species-area

slopes across studies, regions, or organisms because of the influence of ancillary factors on

the relationship.

Fish species richness and composition in Northern Wisconsin lakes (ranging in size

from <1 to 1600 ha) are influenced by physical attributes such as vegetation diversity,

isolation, winter oxygen levels, and pH (Tonn and Magnuson 1982, Tonn et al. 1983, Rahel

and Magnuson 1983, Rahell984, Rahell986). For Umbra-cyprinid assemblages

(minnows), lake surface area and winter oxygen levels accounted for most of the variation in

species richness across lakes (Tonn and Magnuson 1982). The influence of low winter

oxygen concentrations on fish species richness, however, depended on whether we

considered summer or winter richness. Summer fish richness tended to be higher in lakes

with low winter oxygen concentration, whereas winter richness tended to be low with low

winter oxygen concentration. We believe a disturbance mechanism is responsible for this

pattern. Low winter oxygen concentration may reduce population levels in the winter so that 149 species interactions are reduced and more species can coexist in the summer when conditions improve. This would be most likely if the species most susceptible to winterkill are the dominant competitors or predators. A decrease in abundance of important predators/competitors would enable less dominant species to flourish. In lakes with centrarchid-Esox assemblages, species richness declined as winter oxygen concentration declined. These species tend to be more sensitive to winterkill, which results from low winter oxygen concentrations unless a stream or lake refuge, from which recolonization could occur, is available. Thus, fish assemblages in the small, forested lakes of northern

Wisconsin can be placed into three categories, depending on winter oxygen concentrations and the availability of stream/lake refuges; 1. high winter oxygen- centrarchid-Esox with largemouth bass as the top predator; 2. low winter oxygen and stream/lake refuges­ centrarchid-Esox with northern pike as the top predator, and 3. low winter oxygen- Umbra­ cyprinid.

The relationship between fish richness and physical lake attributes was expanded to include oxygen and pH for lakes along a succ~ssional gradient (Rahel 1986). Three distinct fish assemblages have been identified that reflect a sequential loss of species along a gradient from oligotrophic seepage lakes to bog lakes. Fish species richness is correlated with lake size and habitat complexity (which tend to be correlated with each other), productivity and pH, and winter oxygen concentration. Richness decreases along these environmental gradients as the chemical environment becomes harsher and near-shore habitat is simplified.

The centrarchid assemblage, consisting primarily of sunfish, bullhead, bass, and perch, has the highest species richness and occurs in lakes over a wide range of pH and with no winter !50 anoxia. These lakes tend to be larger in size and have more areas of firm substrate and vegetation. The cyprinid assemblage, consisting of minnows such as dace, shiners, and darters, tends to occur in lakes with low winter oxygen concentrations and pH> 5.2-5.4.

These species are probably excluded from lakes with high winter oxygen because of high predation pressure by piscivores. The Umbra-perca assemblage has the fewest species, consisting of the central mudminnow and yellow perch. This assemblage is found in lakes with low pH and low winter oxygen concentration. These two species are also found in each of the other assemblages; the distinction of this assemblage being that they are the only species tolerant of these harsh conditions. Thus, fish species richness and composition at this scale is controlled by local processes that determine the harshness of the chemical conditions and individual species tolerances to the abiotic and biotic environment.

Primary productivity is an important determinant of species richness in lake ecosystems. Species richness for lacustrine phytoplankton, rotifers, cladocerans, copepods,

macrophytes, and fishes exhibits a hump-shaped pattern with increasing primary

productivity. That is, species richness is low at low productivity, high at intermediate

productivity, and low again at high productivity (Fig. 5). We fitted multiple regression

models relating species richness to both productivity and lake area for 33 well-studied lakes

throughout the world (Dodson eta!. in review). Species richness increased with lake surface

area for rotifers, cladocerans, macrophytes, and fish, but not for phytoplankton or copepods.

Lake surface area had no effect on the relationship between species richness and productivity

for most taxa, except for fishes and phytoplankters, where the species richness-productivity

relationship switched from concave-down to concave-up across a gradient of lake size (Fig. 151 6). For lakes small lakes, < 1000 hectares, phytoplankton richness was highest richness at intermediate productivity. In larger lakes the relationship switched such that the highest richness occurred at low and high productivity levels. For fish in lakes larger than 100 hectares, richness was greatest at intermediate productivity levels. In lakes smaller than 100 hectares, the relationship ranged from being nearly linear to having highest fish richness at low and high productivity levels (Fig. 6). Especially for these two groups of organisms, these results emphasize the importance of considering lake surface area when assessing richness-productivity relationships.

When we compared the results of our spatial survey with lakes where productivity had been experimentally manipulated, we found a variety of richness-productivity relationships. For phytoplankton, no relationship existed between richness and productivity across four lakes in a short-term (four year) experiment investigating the effects of nutrient addition and food web structure (Carpenter et al. 1996). In whole-lake experiments where nutrient concentrations were manipulated for more than a decade (Findlay eta!. 1994) phytoplankton richness decreased with increasing productivity in one lake (L226) and in the other lake (L227) richness was highest at intermediate productivity levels. These different responses may be the result of transient effects, lags in response, and possible shifts to new ecosystem states as nutrient levels (and primary productivity) were altered. For example, the negative relationship in L226 resulted because phytoplankton richness initially increased as

productivity increased, but when nutrient additions stopped and productivity declined,

richness remained high. New immigrants tended to remain in the species pool, even after

nutrient additions were stopped. It is uncertain whether these species will continue to persist 152 or eventually disappear- i.e. is there a lag in response or a shift to a new community state?

These results emphasize the importance of considering temporal scale when assessing species richness-productivity relationships. The richness-productivity relationship may be obscured if lakes are not in equilibrium.

Broad-scale patterns across lakes in a region may not hold at an individual lake level because of unique biotic or abiotic attributes that over-ride the general pattern. Individual lake characteristics appear to influence the relationship between productivity and crustacean zooplankton. In the experimental lakes, we found that crustacean zooplankton richness decreased with increasing primary productivity. This likely resulted from a combination of factors, including competitive interactions, food limitation as phytoplankton species composition changed to more inedible forms, and changes in the chemical environment (low oxygen concentrations and high pH) as productivity increased. For example, in the short­ term experiments, shifts in food size toward larger phytoplankton probably enabled the large herbivore, Daphnia to outcompete smaller zooplankton, thus reducing species richness at high levels of productivity. In the long-term experiment, high pH resulting from the nutrient addition and low oxygen resulting from decomposition of an increased phytoplankton biomass eliminated sensitive zooplankton species, resulting in low richness at high levels of primary production. Generalized relationships between species richness and lake

characteristics may breakdown when idiosyncrasies of individual lakes are considered.

Strong relationships between species richness and characteristics of individual lakes

suggest that species composition may be primarily determined by local biotic and abiotic

conditions and that regional processes are not influential. The notion is that species not !53 occurring in a lake, but common in a region are not found in a lake because they cannot survive the local biotic and abiotic conditions, not because they are unable to cross dispersal barriers. We experimentally tested if the crustacean zooplankton assemblage in Little Rock

Lake was saturated with species by augmenting colonization by a suite of zooplankters common in nearby lakes (Lukaszewski et al. in review). We suspended 300-L enclosures in

Little Rock Lake and filled them with filtered lake water and zooplankton from Little Rock

Lake and other LTER lakes. At the end of the experiment, we found that species richness of experimental enclosures that received new colonists was not significantly higher than control enclosures that did not receive colonists. This suggests that local biotic and abiotic conditions determine which species are able to occur in each lake. Lakes are probably not limited by dispersal of zooplankton propagules; most species are capable of arriving at all lakes, although not all species are capable of surviving and reproducing once they arrive.

The relative importance of local and regional factors in determining species richness is ultimately dependent on the scale of observation. Comparisons of fish assemblages in from Finland and north Wisconsin suggest that the composition of fish assemblages in these two regions can be considered to having been processed through a series of filters that operate on several temporal and spatial scales (Tonn et al. 1990, Figure 7). At the coarsest level, regional mechanisms such as Pleistocene events, dispersal barriers, climatic differences, and geomorphology determine the regional species pool. Fishes that pass through this first filter are then subjected to the next filter which is characterized by habitat attributes of the lake-type (e.g. small, forested lakes). Finally, fine-scale, local processes such as lake size, isolation, habitat complexity, and abiotic and biotic conditions determine 154 fish assemblage composition within individual lakes. Clearly, we can not say whether local or regional processes have a greater influence on biotic structure; both play an influential role at different spatial and temporal scales.

V. Detecting Invasions and Extinctions:

One of the largest obstacles in identifying species that have invaded or gone extinct is that rare species are not always detected in regular sampling procedures. This is particularly a problem for detecting cryptic invaders i.e. species that are native to the region, but new to a particular lake. A rare species that is occasionally detected is difficult to distinguish from a recent invader. A multi-criteria approach for detecting invading species and extinctions was originally applied to five of the LTER lakes (Cisneros 1993). This analysis has recently been expanded include 16 years of data. For each lake, a plot of the number of years each fish species was detected versus the first year it was detected (an invasion plot), enabled the identification of species with strong, weak, or no evidence of invasion (Fig. Sa). The diagonal line represents the maximum number of years that a species could be present for each year of first appearance. A point close to the diagonal line corresponds to a species with high persistence. Extinction graphs were plotted in a similar way (Fig. Sb ). A plot of the number of years a fish species was present versus the last year it was detected provided evidence of extinction. These results were coupled with expected patterns of temporal change in abundance, size range, and within-lake dispersion (percentage of sample sites occupied) to assess the likelihood of an invasion or extinction. Positive trends in each of these areas suggest an invasion, whereas negative trends indicate a possible extinction. !55 Evidence was strong for the invasion of two species; rainbow smelt, an exotic species in Crystal Lake and lake trout, a native species in Trout Lake. That is, both species had a high persistence since they were first detected and exhibited an increase in abundance and range size, and dispersion increased for the smelt (not applicable to Lake Trout). In Crystal

Lake, bluegill, a species native to the region, had strong evidence for invasion based on invasion plots but weak evidence based on trends in abundance, size, and dispersion.

Largemouth bass appears to have invaded Sparkling Lake in 1991, based on strong evidence in invasion plots and abundance trends, but weak size and dispersion trends. The shorthead redhorse appears to have invaded Allequash Lake in 1993. It was absent from the record for the 12 years previous to 1993, but has been detected every year since. Its mean length and abundance have increased since we first detected it, although its dispersion within the lake has remained fairly low. Similarly, black crappie appears to have invaded Big Muskellunge

Lake since 1992. Its dispersion and abundance have increase since 1992, although its size range has exhibited no trend.

No species exhibited strong evidence for extinction in either the extinction plots or the trend analyses. However, the analyses predicted a number of potential future extinctions, based on strong negative trends in abundance, range size and dispersion: pumpkinseed, log perch, and yellow bullhead in Big Muskellunge Lake, and yellow perch and mottled sculpin in Trout Lake. Strong patterns in the extinction plots indicate that the Iowa darter in Big

Muskellunge Lake, cisco in Sparkling Lake, and the blackchin shiner and the golden shiner

in Trout Lake have gone extinct. In fact, the blackchin shiner has not been detected in Trout

Lake for the seven years since 1989 and the golden shiner has not been detected since 1987. !56 Although these methods are useful in identifying the invasion and extinction of some species, results were sometimes inconsistent with the known history of the lakes. We know that rainbow smelt has invaded Sparkling Lake, but the evidence based on invasion plots and trends was not strong because it was first observed in the second year of our sampling, thus . we have little to say about when it actually arrived. Lake trout was detected in Trout Lake as an invading species, but this is likely an artifact resulting from historical stocking then recovery of a rare population. In addition, yellow perch in Trout Lake, appeared to be on the verge of local extinction in 1993, but experienced extremely large recruitment episodes in subsequent years (LTER unpublished data). Despite these shortcomings, these methods represent an improvement over previous attempts to dete1mine extinctions and invasions.

The challenge has always been to distinguish an extinction event from a period of low population abundance and likewise, an invasion from a sudden increase in population density of a rare species.

From these patterns and trends, we estimated turnover rates of fishes in the LTER lakes in the Northern Highland Lake District. Overall, the 11% of the fish species changed per decade. Annual turnover rates ranged from 0 to 1.09% among lakes: Allequash = 0.10%;

Trout Lake= 0.54%; Big Muskellunge = 0.46%; Sparkling Lake = 0.41 %; and Crystal Lake

= 1.09%. These estimates are slightly higher than those obtained by Magnuson et al. (1994) using a method based on changes in richness (0-1.09% ). Because separating out rare species from cryptic invasions of short duration is virtually impossible, these estimates are likely

under-estimates. Nevertheless, these results indicate that invasion and extinction are

important and common processes determining species composition in lake islands. !57

VI. Relative importance of invasion and extinction

As islands in a sea of land, the assemblages of species in lakes are controlled by immigration and extinction events. In previous work, the importance of extinction variables has been argued to be most important in structuring fish assemblages based on broadscale patterns and correlations (Magnuson 1988, Tonn eta!. 1990). This hypothesis was tested by comparing the relative importance of extinction and isolation variables in determining fish species richness in lakes in two lake districts; 1) the Northern Highland Lake District in northern

Wisconsin and the Upper Peninsula of Michigan and 2) Finnish lakes (Magnuson et a!.

1998). The extinction variables considered were related to variables important in structuring fish assemblages in northern Wisconsin lakes: lake size, conductivity (a surrogate for productivity), pH, and maximum lake depth (an indication of oxygen concentration). Several isolation variables were considered that were expected to influence the movement of fishes across the lake landscape: 1) the distance over land from one lake to the nearest surface water connection, 2) the vertical distance over land between lakes (summing up and down over ridges and depressions), 3) the distance along a water course from the study lake to the nearest downstream lake, 4) stream gradient (the average slope along a watercourse from one lake to another), 5) area of nearest lake that would serve as a source pool for new species, and 6) the distance to the nearest road.

These analyses were somewhat complicated by the fact that many of the isolation and extinction parameters are correlated. Some correlates result from limnological or geomorphic principles, such as the relation between pH and conductivity or between vertical !58 land distance and horizontal land distance. In addition, road distance and pH are negatively correlated in Wisconsin because small, acidic lakes tend to be surrounded by sphagnum wetlands and therefore road access is limited. One of the most important determinants in correlations among variables is the landscape position of lakes in Wisconsin. Lakes high in the landscape tend to be smaller, more acidic, less fertile, and more isolated than lakes low in the landscape (Kratz eta!. 1997). The multivariate approach used for these analyses was especially useful in teasing apart the correlations to determine the relative importance of the

10 explanatory variables (Magnuson et aL 1998).

Extinction variables usually predicted fish species richness more effectively than did isolation variables in both Wisconsin and Finland lakes (Magnuson et a!. 1998). This does not suggest that isolation is unimportant. Rather, it suggests that differences in the probabilities of immigration and extinction determine their relative importance. Invasion or re-colonization events happen at an extremely low frequency, especially in small, forested lakes with relatively little human activity. In the Wisconsin database, there were several lakes that were fishless despite physical and chemical properties that suggest that they could potentially support fishes. Fishes still have not colonized some lakes, ten thousand years after their formation. In contrast, extinction factors appear to operate over shorter time scales ranging from months to decades. Magnuson eta!. ( 1998) suggest that "the greater the lag in arrival after extinction verses the lag in extinction after arrival, the more important extinction variables will be relative to isolation variables". Therefore, in isolated lakes, extinction factors will appear more important than isolation factors. !59 Aspects of the geomorphology of the two regions suggest that connectivity is important in determining richness and composition of fish communities (Magnuson et a!.

1998). Although both regions are forested landscapes with high densities of lakes, the

Wisconsin region has lower topographical relief and therefore lakes tend to be connected by

groundwater flows and have few stream connections. In contrast, the Finnish lakes are

situated on a landscape with more topographical relief and are highly connected via streams.

Thus, richness and community composition in Finnish lakes was best predicted by stream

gradient and area of the closest connected lake (related to size of the source species pool),

whereas in Wisconsin distance between lakes was a better predictor.

Lake surface area was related to the likelihood of extinction in both regions. In

northern Wisconsin lakes, pH was also important, whereas conductivity was important in

Finnish lakes. These differences too, appear to be influenced by the geomorphic setting of

each area. Two of the extinction variables in Wisconsin lakes are related to the position of

the lake in the landscape (see chapter 3). Lakes high in the landscape tend to have few fish

species and have low conductivity and pH (Kratz eta!. 1997). We have not formally

compared the importance of landscape position between the two areas; these analyses are

underway. Understanding how landscape ecology influences dynamics within individual

lakes is an important challenge currently being undertaken by researchers at NTL-LTER.

VII. Conclusion:

In our studies of lakes as islands, we have emphasized the dynamic nature of aquatic

communities. Temporal changes in species composition of lakes appear to be linked to both 160 within-lake drivers that determine extinction probabilities and landscape-level factors that influence the movement of organisms across the lake district. Detecting changes in species composition has proven difficult because species can be missed when communities are sampled. Our analytical efforts have permitted us to detect invasions and extinctions against the backgroud of sampling problems. These methods have allowed us to estimate realistic species turnover rates. We have begun to understand species dynamics in lake ecosystems as a function of both regional, as well as local processes.

References

Amott, S. E., J. J. Magnuson and N.D. Yan. 1998. Crustacean zooplankton species richness: single- and multiple-year estimates. Canadian Journal of Fisheries and Aquatic Sciences 55:1573-1582.

Arnott, S. E., N.D. Yan, J. J. Magnuson and T. Frost, M. Inter-annual variability and species turnover of crustacean zooplankton in shield lakes. Canadian Journal of Fisheries and Aquatic Sciences, In Press.

Barbour, C. D. and J. H. Brown. 1974. Fish species diversity in lakes. The American Naturalist 108:473-489.

Browne, R. A. 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. Journal of Biogeography 8:75- 83.

Carpenter, S. R., J. F. Kitchell, K. L. Cottingham, D. E. Schindler, D. L. Christensen, D. M. Post and N. Voichick. 1996. Chlorophyll variability, nutrient input, and grazing: evidence from whole-lake experiments. Ecology 77:725-735.

Cisneros, R. 0. 1993. Detection of cryptic invasions and local extinctions of fishes using a long-term database. Thesis. University of Wisconsin-Madison, Madison, Wisconsin.

Diamond, J. M. 1969. Avifauna! equilibria and species turnover rates on the Channel Islands of California. Proceedings of the National Academy of Science 64:57-63. 161 Dodson, S. 1992. Predicting crustacean zooplankton species richness. Limnology and Oceanography 37:848-856.

Eadie, J. M., A. Hurly, R. D. Montgomerie and K. L. Teather. 1986. Lakes and rivers as islands: species-area relationships in the fish faunas of Ontario. Environmental Biology of Fishes 15:81-89.

Eadie, J. M. and A. Keast. 1984. Resource heterogeneity and fish species diversity in lakes. Canadian Journal of Zoology 62: 1689-1695.

Findlay, D. L., R. E. Hecky, L. L. Hendzel, M.P. Stainton and G. W. Regehr. 1994. Relationship between N -fixation and heterocyst abundance and its relevance to the 2 nitrogen budget of Lake 227. Canadian Journal of Fisheries and Aquatic Sciences 51:2254-2266.

Gotelli, N.J. and W. G. Kelley. 1993. A general model ofmetapopulation dynamics. Oikos 68:36-44.

Jenkins, D. G. 1995. Dispersal-limited zooplankton distribution and community composition in new ponds. Hydrobiologia 313/314:15-20.

Keddy, P. A. 1976. Lakes as islands: the distributional ecology of two aquatic plants, Lemna minor L. and L. Trisulca L. Ecology 57:353-359.

Kratz, T. K., K. E. Webster, C. J. Bowser, J. J. Magnuson and B. J. Benson. 1997. The influence of landscape position on lakes in northern Wisconsin. Freshwater Biology 37:209-217.

Lassen, H. H. 1975. The diversity of freshwater snails in view of the equilibrium theory of island biogeography. Oecologia 19:1-8.

MacArthur, R. H. and E. 0. Wilson. 1967. The Theory of Island Biogeography. Princeton University Press, Princeton.

Magnuson, J. J. 1976. Managing with exotics-a game of chance. Transactions of the American Fisheries Society 105:1-9.

Magnuson, J. J. 1988. Two worlds for fish recruitment: Lakes and oceans. American Fisheries Society Symposium 5:1-6.

Magnuson, J. J., B. J. Benson and A. S. McLain. 1994. Insights on species richness and turnover from long-term ecological research: fishes in north temperate lakes. American Zoologist 34:437-451. !62

Magnuson, J. J. and R. C. Lathrop. 1992. Historical changes in the fish community. 553 553 in J. F. Kitchell, J. F. Kitchell. Food Web Management: a case study of Lake Mendota. Springer-Verlag, New York.

Magnuson, J. J., T. W. M., A. Banerjee, J. Toivonen, 0. Sanchez and M. Rask. 1998. Isolation versus extinction in the assembly of fishes in small northern lakes. Ecology 67:

Maguire Jr., B. 1963. The passive dispersal of small aquatic organisms and their colonization of isolated bodies of water. Ecological Monographs 33: 161-185.

Minns, C. K. 1989. Factors affecting fish species richness in Ontario lakes. Transactions of the American Fisheries Society 118:533-545.

Proctor, V. W. and C. R. Malone. 1965. Further evidence of the passive dispersal of small aquatic organisms via the intestinal tract of birds. Ecology 46:728-729.

Rahel, F. J. 1986. Biogeographic influences on fish species composition of northern Wisconsin lakes with applications for lake acidfication studies. Canadian Journal of Fisheries and Aquatic Sciences 43:124-134.

Rahel, F. J. and J. J. Magnuson. 1983. Low pH and the absence of fish species in naturally acidic Wisconsin lakes: Inferences for cultural acidification. Canadian Journal of Fisheries and Aquatic Sciences 40:3-9.

Tonn, W. M. and J. J. Magnuson. 1982. Patterns in the species composition and richness of fish assemblages in northern Wisconsin lakes. Ecology 63:1149-1166.

Tonn, W. M., J. J. Magnuson and A.M. Forbes. 1983. Community analysis in fishery management: An applicaiton with northern Wisconsin lakes. Transactions of the American Fisheries Society 112:368-377.

Tonn, W. M., J. J. Magnuson, M. Rask and J. Toivonen. 1990. Intercontinental comparison of small-lake fish assemblages: the balance between local and regional processes. The American Naturalist 136:345-75. 163 Figure Captions

Figure 1: Northern Highland lakes, Wisconsin.

Figure 2: Influence of the number of years between sampling on estimates of species

turnover averaged across all 7 lakes. Panel A is for all species, panel B is with rare

species removed. The dashed line indicates the turnover between adjacent years

unadjusted by t. Taken from Magnuson et a!. 1994.

Figure 3: The relationship between apparent turnover rates and sampling turnover rates for

phytoplankton from the ELA, crustacean zooplankton from the Dorset lakes, and fishes

from the NTL-LTER. Each point represents the mean annual turnover rates for a lake.

The dashed line indicates where apparent turnover equals sampling turnover. The

difference between this line and points higher than the line provide a conservative

estimate of species turnover attributable t? actual immigration and extinction of species.

Points below the line indicate lakes where sampling error was too high to provide an

estimate of actual turnover.

Figure 4: Asymptotic species richness plotted against mean annual species richness for

phytoplankton, crustacean zooplankton, and fishes. The lower dashed line indicates

where asymptotic richness equals mean annual richness. The upper dashed line indicates

where the asymptotic richness is twice the mean annual richness. 164

Figure 5: A regression analysis of log( species richness) as a function of log(primary

productivity), with the fitted quadratic model assuming no effect of area on the richness­

productivity relationship. Models for all taxa but fish are statistically significant. From

Dodson et a!. in review.

Figure 6: Response surfaces for the interactive effects of lake area and primary productivity

on the species richness of phytoplankton and fishes. Each curve indicates the relationship

between log( richness) and log (primary productivity) for a particular lake size, in

hectares. To prevent extrapolation beyond the range of available data, curves are drawn

only for the region of primary productivity for which we actually had data on species

richness. From Dodson et a!. in review.

Figure 7: Conceptual model of the origin and maintenance of fish assemblages in small

forest lakes, illustrating the effects of filters operating on faunal characteristics and

community structure on different spatial and temporal scales. From Tonn eta!. 1990.

Figure 8: Panel A is a graph to detect invasions in a long-term record of presence/absence

data based on the persistence of a species when present and the first year it was seen.

Panel B is a conceptual graph to detect extinctions in a long-term record of

presence/absence data. From Cisneros 1993. 165

Fig. 1 166

A. All Species 20

~ 15 "'> 0 Percent Turnover c ~ ::> ~ 100(1•EY(S1•S

0 2 3 4 5 6 Years between Sampling

B. Rare Species Removed 15

~., > 0 10 c L-==--=----~---········--· ······--··········-·-·············-·················--·-·············· ~ Percent Turnover ::> 1- ~ 100(1+EY(S1+S2) c - ~ 100(1<-EY(Sl+S<~ 0"' 5 ~., --<.'---- 9.61\ a.

0 2 3 4 5 6 Years between Sampling

Fig. 2 !67

"'-'<; • 40 ~v./ ,_ ,.:,. .·~ . (].) '>- •• •• ... phytoplankton ...... ' 6 30 ·· ,_c ...·· .· fish ::J \D D ... ··~···· ': 20 c ..!tilh .·u ,_(].) l,!jt:J ...... •• m 0..10 Q.. •• <( ...... - '-zooplankton o+----.-----.r----.-----.--.. ·· 0 10 20 30 40 Sampling Turnover

Fig. 3 168

~ j ..-< 0..

0 • 0 •• ...-< ·•·•··•··•... •• \ •• •• • 0 0 •• 00 ··.IS\~ 0 •• •• •• •••••• •••• 0 ~·..p·. \0 u .. • ~P:·· ~.·. /:•... ~·. ·~·.(... •• ·.• ~Cl'l •• • .. . .,.....( .8: ·..

0 Cl:l • • N ..-< • 0.. 0. ·. 0o...... -... _. ••• •• •• N ··: 0 0 0 0 0 0 0 tn N ...-< 169 Phytoplankton Copepods ••• • R2 = 0.51, p < 0.001 2 .. • • • • 1 1 • • 0 0 R2 = 0.40, p = 0.01 - 0 1 2 3 4 0 1 2 3 4 ..-.. 1/) Rotifers Macrophytes 1/) 2 2 (l) R = 0.54, p < 0.001 R = 0.46, p = 0.003 c .c 2 2 • (.) • ·;: • • 1/) • (l) 1 • • •• 1 , . "(3 (l) • • 0. • • 1/) • 0 • • 0 -.....0 C) 0 0 1 2 3 4 0 1 2 3 4 Cladocerans Fish R2 = 0.49, p < 0.001 R2 = 0.48, p < 0.001 2 2 • • '·• 1 1 •• • • • • • • • • 0 • • 0 • • 0 1 2 3 4 0 1 2 3 4

log10(PPR)

Fig. 5 170 Phytoplankton 2.5 1Q5

2.0 4 101 10

'10'1 - 1.5 Ill Ill (1) 10° c .c ....(.) Ill (1) 1 2 3 ·~ Fish ~ 2.0.--.--.--.--~~--~~ .!!!.- ,...0 O'l 0 1.5

1.0

0.5

OL-L...:..J.._ __.r___t_ __ ..L._____L __ _l___J 0 1 2 3

Fig. 6 171

• Pleistocene eventa Regional • Dispersal barriers Processes • Climatic differences • Geomorphlc/edaphlc limits

Regional Species Pool !~~a.no

Lake-Type • Abiotic conditions Characteristics • Resource distribution • Habitat stability & complexity

Small-Lake Species Pool e.g., small lakes In northern Wisconsin

Local Processes • Area • Structural complexity • Isolation • Abiotic conditions • Biotic Interactions

e.g., Jude Lake

Fig. 7 172 A INVASION PLOT p N E u M R B '-...---.-Maximum possible s E I R persistence s 0 T F E N y c E E A R s

FIRST SAMPLING YEAR LAST FIRST YEAR SEEN B N EXTINCTION PLOT p u E M R B E s R Maximum possible I persistence s 0 T F E y N E c A E R s

FIRST SAMPLING YEAR LAST LAST YEAR SEEN [J Early absence/ some evidence

Fig. 8 • Intermediate absence/ some evidence

!il1Jt,~;H:~ Late absence/ no ev1"dence 173 SUMMARY OF THESIS

Communities in north temperate lakes appear dynamic with species appearing and disappearing from the sampling record every year. These appearances and disappearances probably result from two processes; 1. Immigration and extinction of species, and 2.

Sampling error caused by failing to detect species that were present. Both components have important implications for how we monitor and manage aquatic systems.

Apparent species turnover rates in north temperate lakes were high, averaging 18% for phytoplankton, 16% for zooplankton, and 20% for fishes. High rates of change were further supported by the decline in community composition similarity between increasing time intervals. Turnover rates varied among years, emphasizing the importance of long-term studies in understanding community dynamics.

High apparent turnover rates (regardless of whether they result from actual turnover or sampling error) have important implications for the assessment of species richness. In general, annual richness estimates ranged from approximately 113 to 2/3 the total species pool, based on 10 to 16 years of sampling. Because apparent turnover rates varied among lakes, the size of the total species pool could not be predicted from single year estimates.

That is, lakes with the lowest mean annual richness sometimes had the highest long-term richness estimates because species composition was highly dynamic even though annual

richness tended to remain constant from year to year. The lack of correlation between long­

term and short-term richness estimates meant that the relationship between richness and

environmental variables also changed with the temporal scale of study. These results suggest

that care must be taken to standardize richness assessments, i.e. they should be used as an 174 index rather than an absolute indication of the number of species in a system. Furthermore, the use of short-term versus long-term richness should depend on the type of question being addressed. Short-term assessments are likely indications of the number of potentially interacting species and therefore are probably best suited for questions considering the influence of environmental change on community structure. Long-term assessments take into account shifts in species composition and may be best suited for studies where the investigator is interested in quantifying the potential species pool. This type of information may be useful, for example, when investigating richness/stability relationships.

Sampling error was high for all three taxonomic groups. Despite consistent methodology and multiple samples per year, we failed to reliably detect all species every year. Using a Monte Carlo simulation model, we determined the potential contribution of sampling error to the apparent turnover rate. For phytoplankton, potential sampling error was lower than apparent turnover in each lake, suggesting that perhaps actual immigration and extinction events were occurring. For zooplankton and fishes, sampling error was generally higher than the apparent turnover. Because of this high sampling error, we cannot be certain if changes result from immigration and extinction of species.

Despite an intensive sampling programs, many species were not reliably detected each year. In fact, on average, less than one third of the long-term species pool was detected every year. Actual turnover rates varied among taxonomic groups in a manner that was consistent with expectations based on life-history traits and dispersal abilities. Phytoplankton have the shortest generation times, shortest lifespans, high dispersal rate, high population densities, and high growth rates; Coincidentally, they also have the highest actual turnover 175 rates (6.2% per year). Zooplankton have life-history traits intermediate between phytoplankton and fishes and likewise, had intermediate actual turnover rates; 0.7%. Fishes, with undetectable actual turnover, have low immigration probabilities, long lifespans, and relatively long generation times.

These results have important implications for monitoring and conservation of biodiversity. With regard to monitoring biodiversity, high turnover rates suggest that species pools are larger than we first anticipate. If we truly wish to know how many species are in a particular habitat then we must intensify our sampling programs to include long-term monitoring. It is important to note that in the analysis of zooplankton richness, a comparison of equal effort of spatial and temporal sampling resulted in similar species richness.

However, the identity of the species was different, indicating that both spatial and temporal expansion of sampling programs have merit.

Long-term monitoring of aquatic communities also revealed that approximately 2/3 of the community was transient; i.e., they were not detected every year. Furthermore, 1/3 to 1/2 of the species was detected in less than 25% of the years. A large proportion of the species in lakes are rare and are probably missed in routine sampling. Although sampling error was highest in this group of species we can not determine with certainty whether their low persistence is the result of not detecting them when they are present or the result of immigration and extinction. Regardless, these rare species may play important roles in

community dynamics. They may serve important functions as 'reserve' species that are able

to quickly respond to changes in environmental conditions caused by climate change,

invasion of new species, etc. On the other hand, they may simply be species that maintain 176 low densities and may not significantly contribute to ecosystem processes. The role of rare species in communities merits further investigation. llllllllllllllllllllllllllllllllllllllllllllllllll/1